fix(composer): close binding_source code injection, line-align, and project_root UnboundLocalError

Three findings from CodeRabbit's review of 6c23994b, all verified against
current code before fixing:

- manager.py.j2 interpolated binding.source unescaped into a Python comment
  (`pass  # dynamic_text binding_source "{{ el.binding_source }}" draws
  nothing`). A source string with a newline broke out of the comment; a
  crafted payload produces a clean, ast.parse-valid `import os` in the
  generated plugin (confirmed against the pre-fix template). This is now a
  fixed literal comment that never interpolates the value. Live now that
  composer_bp is registered. CWE-94.
- _alignElement moved a line's x0 (or y0) to the new position but left x1
  (or y1) behind, so aligning a line changed its shape instead of moving
  it. Both endpoints now translate by the same delta.
- web_interface/app.py only assigned project_root inside the relative-path
  branch of the plugins_dir resolution. An absolute plugin_system.plugins_
  directory (a supported config value) hit UnboundLocalError importing the
  module at all, since SchemaManager/composer_bp use project_root further
  down. Now assigned unconditionally before the branch.

Also extends BOUND_TYPES coverage in composer-app.js (_isBound,
removeConfigVar, _validateBeforeExport) from dynamic_text/progress_bar to
all six element types that carry a binding object (countdown, pips,
sparkline, gauge too) -- found by direct code reading against
ELEMENT_DEFAULTS in composer-canvas.js, not from a review comment. Without
it, those four types could export with an unbound config key with no
validation error, and deleting a config var they used gave no warning.

All four fixes have mutation-checked regression tests (fail against the
reverted code, pass with the fix): test_binding_source_cannot_break_out_of_the_comment_it_lands_in,
test_align_translates_both_line_endpoints_not_just_the_start,
test_app_plugins_dir_resolution.py, test_binding_checks_cover_every_bound_element_type.

Full suite: 4365 passed, 58 skipped, 2 failed -- both the pre-existing
Europe/Kiev/Asia/Calcutta tzdata-alias gap on this sandbox, identical on
origin/main, unrelated to this change.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

.github/workflows/test.yml

@@ -72,4 +72,4 @@ jobs:
             --ignore=test/plugins \
             --cov=src --cov=web_interface \
             --cov-report=term \
-            --cov-fail-under=45
+            --cov-fail-under=52

CHANGELOG.md

@@ -17,6 +17,81 @@ release that ships it.
 accepts both, but the store flags the old spelling as deprecated
 (`store_manager.py`) and only the new one is in `schema/manifest_schema.json`.
 
+## 3.3.0
+
+**The release the sports scoreboards floor on to delete their bundled copies.**
+3.2.0 shipped the unified sports library and made `ledmatrix_min_version`
+enforceable; this ships the last three shared modules and completes the store
+gate, so a scoreboard can now floor here and carry no fallback at all.
+
+New modules a plugin may import via `src.*` and floor on 3.3.0 for:
+
+- `src/common/sports_card.py` — settings, colour, font and date helpers for a
+  scoreboard's `game_renderer.py`. Free functions taking `config`/`fonts`
+  explicitly, so nothing about the caller's class is assumed.
+- `src/common/sports_game_renderer.py` — `SportsGameRendererMixin`: scroll/Vegas
+  card geometry (centre gap, logo slot and cache key, layout offsets, the
+  upcoming-card date and time layout). No `__init__` and no state, so adoption
+  is one line on the class statement.
+- `src/common/sports_shared.py` — `SportsCoreSharedMixin`,
+  `SportsLiveSharedMixin`, `SportsRecentSharedMixin`: the `sports.py` bodies
+  byte-identical in all eight lineage-sharing scoreboards.
+
+All three sit under `src/common/` rather than `src/base_classes/sports/`,
+deliberately: importing that package pulls `core.py` → `DisplayManager` →
+`rgbmatrix`, and these are pure logic. Plugins importing them must not acquire a
+hardware dependency.
+
+Three notes for anyone adopting `sports_shared`:
+
+- `SportsRecentSharedMixin` defines `__init__`. Its bare `super()` binds to the
+  mixin, so it reaches the host only when the mixin is listed **first** in the
+  bases. Reversing that order silently skips the host constructor.
+- Methods that resolve the plugin's `config_schema.json` use `_plugin_dir()`,
+  which walks the MRO rather than reading `__file__` — `__file__` is now
+  `src/common/`. It walks because `SportsCore` is an ABC: a subclass built with
+  `type(name, bases, ns)` reports `__module__` as `"abc"`.
+- `_get_timezone`, `_extract_game_details` and `_fetch_data` are byte-identical
+  across the eight but stay in the plugins. The first binds a per-plugin
+  timezone module whose contents differ; the other two are the abstract stubs
+  that define the sport.
+
+**The store's compatibility gate is now on every registry-managed route.** 3.2.0
+gated `install_plugin`. This release gates the git-pull update path and
+`install_from_url`, so a plugin whose floor the core cannot meet is refused
+after download with no partial directory left behind.
+
+### Fixed
+
+- **ESPN 403s.** `site.api` began rejecting the User-Agent strings this repo
+  sent on 2026-08-04; every shared-data-source scoreboard returned
+  `403 Forbidden`. Requests now send an identifying token with a project URL —
+  browser-style strings and bare custom tokens are both refused.
+- **Low-memory boards becoming unreachable under load** while still answering
+  pings and serving the web UI. Fetched payloads are released after delivery
+  rather than pinned on the completed request for up to an hour, malloc arenas
+  are capped, and log volume and SD writes are reduced. Available memory is now
+  reported in Tools diagnostics.
+- **A failed logo download pinning a team to a grey box**: the placeholder was
+  written under the real logo's filename, so later attempts found a file and
+  reported success without retrying.
+- **A plugin enabled but never loaded is retried** rather than staying absent
+  with `error = null`.
+- `ttl` now controls cache expiry; abandoned cache writes no longer leave temp
+  files; one cache-cleanup thread per directory rather than per manager.
+- Odds are fetched for the games displayed, not the whole schedule window, and a
+  stalled ESPN no longer stalls the whole plugin update.
+- Array-item secrets are no longer wiped or logged.
+- A restored backup matches the device it was taken from.
+- The web UI reports the real error instead of "unknown", rejects non-finite
+  JSON numbers, and stops checkbox groups posting back hidden options.
+
+### Added
+
+- `display.hardware.orientation` for panels mounted upside down.
+- Vegas keeps live content in the ticker rather than being preempted by it.
+- Schemas can label enum dropdown options.
+
 ## 3.2.0
 
 **The first release shipping the unified sports library.** This is the version

README.md

@@ -161,9 +161,11 @@ The system supports live, recent, and upcoming game information for multiple spo
 ### LED Matrix Panels  
 (2x in a horizontal chain is recommended)
 - [Adafruit 64×32](https://www.adafruit.com/product/2278) – designed for 128×32 but works with dynamic scaling on many displays (pixel pitch is user preference)
+**Warning: Lately the Waveshare Panels have had different variations - only some are compatible with this project. I hope to identify what is different to fix it but so far there is a decent chance you get a mis-matched set of panels if you don't buy them all at once! **
 - [Waveshare 64×32](https://amzn.to/3Kw55jK) - Does not require E addressable pad
 - [Waveshare 96×48](https://amzn.to/4bydNcv) – higher resolution, requires soldering the **E addressable pad** on the [Adafruit RGB Bonnet](https://www.adafruit.com/product/3211) to “8” **OR** toggling the DIP switch on the Adafruit Triple LED Matrix Bonnet *(no soldering required!)*
-  > Amazon Affiliate Link – ChuckBuilds receives a small commission on purchases  
+- There are some Panels on Aliexpress that have worked fine for me, shop around! I think Adafruit is probably the "safest" but they do have some limitation on resolution and layout.
+  > Amazon Affiliate Links – ChuckBuilds receives a small commission on purchases  
 
 ### Power Supply
 - [5V 4A DC Power Supply](https://www.adafruit.com/product/658) (good for 2 -3 displays, depending on brightness and pixel density, you'll need higher amperage for more)
@@ -600,6 +602,14 @@ These settings are typically only needed for non-standard panels or custom confi
   - Leave empty unless you need custom mapping
   - See rpi-rgb-led-matrix documentation for full options
 
+- **`orientation`** (string, default: "normal")
+  - Rotates the rendered image to match how the panel is physically mounted
+  - Set to `"180"` (or use the "Upside Down" option in the web UI's Display
+    settings) if the panel is mounted upside down — useful for optimizing
+    where the Raspberry Pi and wiring sit relative to the mounting location
+  - Applied independently of `pixel_mapper_config` (appended as a trailing
+    `Rotate:180` mapper), so custom mapper configs keep working alongside it
+
 - **`row_address_type`** (integer, default: 0)
   - How rows are addressed on the panel
   - Most panels use 0 (direct addressing)

config/config.template.json

@@ -112,6 +112,7 @@
             "led_rgb_sequence": "RGB",
             "limit_refresh_rate_hz": 100,
             "pixel_mapper_config": "",
+            "orientation": "normal",
             "row_address_type": 0,
             "multiplexing": 0,
             "panel_type": ""
@@ -129,6 +130,9 @@
         "plugin_rotation_order": [],
         "use_short_date_format": true,
         "vegas_scroll": {
+      "live_in_ticker": false,
+      "live_weight": 3,
+      "favorite_live_weight": 5,
             "enabled": false,
             "scroll_speed": 50,
             "separator_width": 32,
@@ -149,7 +153,7 @@
             "min_plugin_width": 8,
             "lead_in_width": 0,
             "plugins_per_cycle": 6,
-            "max_plugin_width_ratio": 3.0,
+            "max_plugin_width_ratio": 0.0,
             "overflow_mode": "rotate",
             "dynamic_duration_enabled": true,
             "min_cycle_duration": 60,

docs/ADVANCED_FEATURES.md

@@ -64,10 +64,98 @@ JSON is optional.
 | `target_fps` | `125` | Target frame rate |
 | `buffer_ahead` | `2` | Number of plugins buffered ahead |
 
-This table is a subset — `display.vegas_scroll` supports 26 keys in
+This table is a subset — `display.vegas_scroll` supports 30 keys in
 total. See the full list in
 [CONFIG_REFERENCE.md](CONFIG_REFERENCE.md#displayvegas_scroll--continuous-scroll-mode).
 
+### Live Content in the Ticker
+
+By default, live content **preempts** Vegas mode: while any plugin reports
+live priority, the display controller refuses to run the ticker and shows
+that plugin's full-screen display instead. You get a big readable scoreboard,
+but the marquee stops entirely for the duration of the game.
+
+Set `live_in_ticker` to keep the ticker running and let live content take
+**extra turns inside it** instead:
+
+```json
+"vegas_scroll": {
+  "live_in_ticker": true,
+  "live_weight": 3,
+  "favorite_live_weight": 5
+}
+```
+
+#### Why weights exist
+
+The rotation is otherwise a strict round robin — every plugin appears exactly
+once per cycle. With a dozen plugins enabled, a live score comes round once a
+lap and can be minutes old by the time you see it. A weight of *N* gives a
+plugin *N* slots per cycle.
+
+The slots are placed by **Smooth Weighted Round-Robin**, the same scheduler
+the sports plugins use internally to rotate their own games. The important
+property is that repeats are *spread through the cycle* rather than clumped:
+three appearances in a row followed by a long silence would be worse than not
+boosting at all.
+
+Twelve plugins, with a favorite's baseball game and an ordinary live hockey
+game (`live_weight: 3`, `favorite_live_weight: 5`):
+
+```
+baseball > hockey > weather  > clock  > baseball
+stocks   > news   > flights  > baseball > hockey
+calendar > f1     > music    > baseball > tides
+birds    > hockey > baseball
+```
+
+18 slots for 12 plugins. Baseball appears 5 times, hockey 3, everything else
+once, and no plugin ever appears twice in a row — **including across the seam**
+where the cycle loops back on itself. Smooth Weighted Round-Robin schedules the
+heaviest item first and usually last as well, so the strip would otherwise show
+it twice running at exactly the one join a within-cycle check cannot see. The
+trailing repeat is moved into the widest remaining gap. Where a double is
+unavoidable — a plugin holding most of the slots has to neighbour itself — the
+schedule is left as it is.
+
+#### Where the weight comes from
+
+For each plugin in the rotation, in order:
+
+1. **The plugin's own answer.** If it implements
+   `get_vegas_priority_weight()` and returns a number, that wins. This is the
+   only route for favorite-team awareness — the core can see *that* a game is
+   live, but not *whose*, so a scoreboard has to say so itself.
+2. **The core's default.** When the plugin returns `None` (the base-class
+   default), a plugin where both `has_live_priority()` and `has_live_content()`
+   are true gets `live_weight`.
+3. **Everything else** gets 1.
+
+Because of step 2, **existing plugins need no changes** — any scoreboard with
+`live_priority` enabled already gets extra turns. Step 1 is opt-in, for
+plugins that want to distinguish a favorite's game from any other live game.
+
+Weights are clamped to 1–10. A weight of 1 is no boost; a weight below 1 would
+drop the plugin from the rotation entirely, which is never what is meant.
+
+#### Things worth knowing
+
+- **Weights are per plugin, not per game.** A scoreboard showing four live
+  games still occupies one slot at a time, rotating its own games within that
+  slot using its own `favorite_live_boost`. This controls how often the
+  *plugin* comes round.
+- **The ticker is zero-sum.** Giving baseball 5 slots does not make the cycle
+  faster; it makes the cycle *longer* and everything else proportionally
+  rarer. If you want live scores sooner in wall-clock terms, pair this with a
+  smaller `plugins_per_cycle`.
+- **Frequency is not freshness.** Each appearance redraws from the plugin's
+  current data (`refresh_updated_plugins()` drops cached content when a
+  plugin's data changes), but how current that data is depends on the
+  plugin's own `live_update_interval`. Showing a stale score five times a lap
+  is no better than showing it once.
+- **Everything still appears.** A boost never starves another plugin out of
+  the cycle; low-weight plugins keep their single slot.
+
 ### Per-Plugin Configuration
 
 Override Vegas behavior for specific plugins:

docs/CONFIG_REFERENCE.md

@@ -18,7 +18,7 @@ tooling against it.
 | `web_display_autostart` | bool, `true` | Whether the web interface service starts with the system | `scripts/utils/start_web_conditionally.py` |
 | `timezone` | string, `"America/New_York"` | IANA timezone for schedules and displays | `ConfigManager.get_timezone()` |
 | `target_fps` | int, `100` | Frame-rate ceiling for plugin rendering | `src/plugin_system/base_plugin.py`, `src/common/sports_scroll.py` |
-| `location` | object | `city` / `state` / `country`, offered to plugins that need a location (weather, etc.) | plugins via merged config |
+| `location` | object | `city` / `state` / `country`. Supplies the **default** for a plugin's own `location_city` / `location_state` / `location_country` setting, so weather, radar and friends follow this device without being configured twice. A value saved on the plugin itself still overrides it. | `SchemaManager.apply_device_location()`, then plugins via merged config |
 
 ## `schedule` — display on/off hours
 
@@ -66,6 +66,7 @@ in `DisplayManager` (`src/display_manager.py`, ~lines 270–295).
 | `led_rgb_sequence` | string, `"RGB"` |
 | `limit_refresh_rate_hz` | int, `100` (code default 90) |
 | `pixel_mapper_config` | string, `""` — e.g. `"U-mapper"` / `"Rotate:90"` |
+| `orientation` | string, `"normal"` — `"180"` rotates the rendered image 180° for panels physically mounted upside down (e.g. to move the Pi/wiring to a more convenient side); composed onto `pixel_mapper_config` as a trailing `Rotate:180` mapper, so it stays independent of any custom `pixel_mapper_config` value |
 | `row_address_type` | int, `0` — non-standard panel row addressing |
 | `multiplexing` | int, `0` — panel multiplexing scheme |
 | `panel_type` | string, `""` — set to `"FM6126A"` or `"FM6127"` for panels needing init |
@@ -103,7 +104,8 @@ logical image to multiple chained physical panels.
 ## `display.vegas_scroll` — continuous scroll mode
 
 Read by `src/vegas_mode/config.py` (`VegasScrollConfig.from_config`). See
-[ADVANCED_FEATURES.md](ADVANCED_FEATURES.md) for behavior details.
+[ADVANCED_FEATURES.md](ADVANCED_FEATURES.md) for behavior details, including
+[live content in the ticker](ADVANCED_FEATURES.md#live-content-in-the-ticker).
 
 | Key | Type / default |
 |---|---|
@@ -127,13 +129,16 @@ Read by `src/vegas_mode/config.py` (`VegasScrollConfig.from_config`). See
 | `min_plugin_width` | int, `8` |
 | `lead_in_width` | int, `0` |
 | `plugins_per_cycle` | int, `6` |
-| `max_plugin_width_ratio` | float, `3.0` |
+| `max_plugin_width_ratio` | float, `0.0` |
 | `overflow_mode` | string, `"rotate"` |
 | `dynamic_duration_enabled` | bool, `true` |
 | `min_cycle_duration` | int, `60` |
 | `max_cycle_duration` | int, `240` |
 | `frame_based_scrolling` | bool, `true` — frame-count-based scroll stepping |
 | `scroll_delay` | float, `0.02` — seconds between scroll updates (~50 FPS) |
+| `live_in_ticker` | bool, `false` — keep scrolling during live games instead of handing the display to a full-screen scoreboard |
+| `live_weight` | int, `3` (1–10) — slots per cycle for a plugin with live content |
+| `favorite_live_weight` | int, `5` (1–10) — slots per cycle when a plugin reports a favorite team is live |
 
 ## `sync` — multi-display synchronization
 

docs/LOW_MEMORY_BOARDS.md

@@ -0,0 +1,115 @@
+# Running on Low-Memory Boards
+
+Applies to the Pi Zero 2 W (512 MB), Pi 3 / 3B+ (1 GB), and the 1 GB Pi 4.
+If your board has 2 GB or more you can skip this document.
+
+## The failure this prevents
+
+The display process is the largest thing on the board. On a 1 GB Pi 3B+ with
+around 20 plugins enabled it settles near **600 MB of 905 MB usable**, leaving
+under 200 MB of headroom for everything else.
+
+When that headroom runs out, the board does not crash cleanly. `fork()` starts
+failing, and because a new process is needed to do almost anything, the
+symptoms look nothing like "out of memory":
+
+| What you see | Why |
+|---|---|
+| SSH accepts the connection then closes it instantly, before any banner | `sshd` forks a session per connection; the fork fails |
+| The web UI still responds quickly | Already running, serves from existing threads, forks nothing |
+| Ping is perfect, 0% loss | Handled entirely in the kernel |
+| The panel is dark | The display process was killed and cannot be respawned |
+| The clock is wrong after the next boot | `fake-hwclock`'s periodic save is a scheduled job, and it cannot fork either |
+
+The board looks healthy from the outside and cannot be logged into. Only a
+power cycle clears it. If you are here because SSH stopped working, also see
+[SSH_UNAVAILABLE_AFTER_INSTALL.md](SSH_UNAVAILABLE_AFTER_INSTALL.md), which
+covers the more common cause (AP mode).
+
+## Check your headroom
+
+```bash
+free -m
+ps -eo rss,comm --sort=-rss | head -5
+```
+
+If `MemAvailable` is under ~150 MB while the display is running, you are close
+to the edge. To watch it over time:
+
+```bash
+watch -n 30 'free -m | head -2'
+```
+
+Available memory that falls steadily rather than holding flat means you will
+reach the wall; it is a question of when.
+
+## What to do
+
+**1. Enable the memory cgroup controller.** Without it, the `MemoryMax=85%` in
+`systemd/ledmatrix.service` is accepted by systemd and silently ignored, so the
+service has no ceiling and a runaway takes the whole board down instead of just
+restarting. Raspberry Pi firmware disables this controller by default.
+
+`first_time_install.sh` does this for you. To check it took effect:
+
+```bash
+grep memory /sys/fs/cgroup/cgroup.controllers
+```
+
+If that prints nothing, add `cgroup_enable=memory cgroup_memory=1` to the
+kernel command line and reboot. Edit whichever file your image uses —
+`/boot/firmware/cmdline.txt` on current Raspberry Pi OS, `/boot/cmdline.txt` on
+older layouts (the installer checks the first and falls back to the second).
+Everything must stay on a single line.
+
+This changes the failure mode from "the board becomes unreachable" to "the
+display service restarts". It is a safety net, not a fix.
+
+**2. Run fewer plugins.** This is the actual remedy. Every enabled plugin costs
+memory permanently — its module, its parsed config, and its cached API
+responses. On a 512 MB or 1 GB board, keep the enabled set small and prefer
+plugins that poll infrequently.
+
+**3. Lower the cache ceiling.** The in-memory cache is sized from total RAM
+(150 entries at 1 GB and below, up to 1500 at 8 GB). To go lower still:
+
+```ini
+# /etc/systemd/system/ledmatrix.service.d/override.conf
+[Service]
+Environment=LEDMATRIX_CACHE_MAX_ENTRIES=75
+```
+
+Writing the file does not change the running service. Reload systemd and
+restart it:
+
+```bash
+sudo systemctl daemon-reload
+sudo systemctl restart ledmatrix
+```
+
+Fewer entries means more API calls, so lower this only while you are actually
+short of memory.
+
+**4. Consider `MemoryHigh`.** `MemoryMax` kills and restarts. `MemoryHigh`
+throttles and reclaims instead, which is gentler — but on a board where the
+process genuinely wants more than the limit, sustained reclaim can stall the
+render loop and show as visible stutter on the panel. Add it only if you prefer
+degraded output to a restart:
+
+```ini
+[Service]
+MemoryHigh=70%
+```
+
+## Keep your logs
+
+These images default to volatile journald storage, so every reboot destroys the
+logs — including the ones explaining why the board rebooted. `first_time_install.sh`
+enables persistent storage capped at 64 MB. To confirm:
+
+```bash
+journalctl --list-boots
+```
+
+More than one boot listed means logs are surviving reboots. If only one is
+listed, journald is still writing to `/run` (tmpfs).

docs/PLUGIN_API_REFERENCE.md

@@ -170,6 +170,47 @@ Default returns `False`.
 List of display modes to show during a live takeover. Default returns the
 plugin's `display_modes` from its manifest.
 
+#### `get_vegas_priority_weight() -> Optional[int]`
+
+How many slots per Vegas cycle this plugin should get. Default returns
+`None`, which defers to the core.
+
+The Vegas ticker is otherwise a strict round robin — every plugin appears
+exactly once per cycle — so with a dozen plugins enabled a live score can be
+minutes stale by the time it comes round. A weight of *N* gives the plugin
+*N* slots per cycle, spread evenly through it rather than clumped.
+
+**You usually do not need this.** When the hook returns `None`, the core
+already gives a plugin `vegas_scroll.live_weight` whenever
+`has_live_priority()` and `has_live_content()` are both true. Live sports get
+extra turns with no code at all.
+
+Implement it only when the plugin knows something the core cannot. The
+motivating case is favorite teams — the core can see *that* a game is live,
+but not *whose*:
+
+```python
+def get_vegas_priority_weight(self):
+    if not (self.has_live_priority() and self.has_live_content()):
+        return None                       # let the core decide
+    vegas = self.global_config.get('display', {}).get('vegas_scroll', {})
+    if self._favorite_is_live():
+        return vegas.get('favorite_live_weight', 5)
+    return vegas.get('live_weight', 3)
+```
+
+The weight is per *plugin*, not per game: a scoreboard showing four live games
+still occupies one slot at a time and rotates its own games within it. Values
+are clamped to 1–10 by the caller. An exception here is caught and logged, and
+the core then falls back to its own live-content check — so a plugin whose
+weight calculation is broken still gets `live_weight` for a game that really
+is live, rather than being demoted to 1.
+
+Only consulted when the user has set `vegas_scroll.live_in_ticker`. With the
+default (`false`) live content preempts Vegas entirely and there is no ticker
+to be weighted within. See
+[ADVANCED_FEATURES.md](ADVANCED_FEATURES.md#live-content-in-the-ticker).
+
 ### Vegas scroll hooks
 
 Vegas mode shows multiple plugins as a single continuous scroll instead of

docs/README.md

@@ -14,6 +14,7 @@ the one-shot installer. The pages here go deeper.
 5. [TROUBLESHOOTING.md](TROUBLESHOOTING.md) — common issues and fixes
 6. [SSH_UNAVAILABLE_AFTER_INSTALL.md](SSH_UNAVAILABLE_AFTER_INSTALL.md) — recovering SSH after install
 7. [CONFIG_DEBUGGING.md](CONFIG_DEBUGGING.md) — diagnosing config problems
+8. [LOW_MEMORY_BOARDS.md](LOW_MEMORY_BOARDS.md) — Pi Zero 2 W / 3B+ / 1GB Pi 4 memory limits
 
 ## I want to write a plugin
 

docs/SPORTS_UNIFICATION.md

@@ -27,7 +27,7 @@ These are independent concerns. Conflating them is what produces god classes.
 | Plugin loads on a core that predates a module | Guarded import with a bundled fallback (`try: from src.X import Y / except ModuleNotFoundError: from y import Y`) |
 | Plugin loads on a core that predates a *method* | Capability probing — `hasattr(SportsCore, "_detect_stale_games")` — never a version comparison. The loader's compat check is advisory-only (it logs and continues), so probing is the real protection. |
 | Core changes never break a plugin's rendering | The **view-model contract**: `_extract_game_details_common` returns a dict whose `GUARANTEED_KEYS` are frozen by `test/test_skin_system.py::TestViewModelContract`. Keys may be added, never renamed or removed. |
-| A plugin can drop its bundled copy safely | The **sunset rule**: its manifest must floor `ledmatrix_min_version` at the first core release shipping the module (recorded in `CHANGELOG.md`) — *necessary but not sufficient*. Nothing enforces that floor today, so the copy also waits for the B6 gate below. |
+| A plugin can drop its bundled copy safely | The **sunset rule**: its manifest must floor `ledmatrix_min_version` at the first core release shipping the module (recorded in `CHANGELOG.md`) — *necessary but not sufficient*. The store enforces that floor on every registry-managed install and on both supported update paths (sideloading via `install_from_url` is not gated), but a floor cannot reach a user who never updates, so the copy also waits for the B6 gate below. |
 
 The core API is **additive-only**. A method the plugins call is never removed or
 given a new required parameter; new behavior arrives as new methods with
@@ -204,7 +204,7 @@ legacy compatibility rather than the mechanism.
 B0–B3 are merged and shipping in core 3.2.0. Everything that remains is
 **rollout**, and it splits into three phases with very different risk profiles.
 The original plan folded the last two together; they are separated here because
-one of them is safe by construction and the other is not.
+one of them cannot break a user on an old core and the other can.
 
 | Phase | Scope | Status | Gate |
 |---|---|---|---|
@@ -212,9 +212,9 @@ one of them is safe by construction and the other is not.
 | **B1** | Promote the nine universal methods; convert `sports.py` → package | ✅ | Characterization suite green; no behavior change intended |
 | **B2** | `CelebrationMixin` + rotation strategies as opt-in capabilities | ✅ | Non-adopters have zero new code in their MRO; strategies checked against verbatim plugin transcriptions |
 | **B3** | Upstream the scroll **orchestration** layer as `src/common/sports_scroll.py`, reading `global_config['target_fps']` natively | ✅ | Content building stays per-sport |
-| **B4** | Ship 3.2.0 *and* make version reporting trustworthy | ⏳ **next** | Tag, release, and `src.__version__` agree; compatibility gate merged |
+| **B4** | Ship 3.2.0 *and* make version reporting trustworthy | ✅ | Released 2026-08-03; tag, release and `src.__version__` agree; compatibility gate merged (#428, #431, #433) |
-| **B5** | Adoption — guarded core imports: three pilots, then the remaining six. **Bundled copies stay.** | after B4 | Per plugin: harness + goldens byte-identical, then a device soak |
+| **B5** | Adoption — guarded core imports, all eight. **Bundled copies stay.** | ✅ | All eight adopted; harness byte-identical; see the B5 retrospective below — four shipped broken and were repaired in plugins #251 |
-| **B6** | Sunset — delete the bundled copies | **blocked** | B4's gate shipped *and* in users' hands (see below) |
+| **B6** | Sunset — delete the bundled copies | ✅ | Ran 2026-09-01, all eight. Floors at 3.2.0; the store refuses on all three routes in (#431/#433, #508, #510). See "B6 — what actually happened" |
 
 ### B4 — what "ship 3.2.0" actually requires
 
@@ -234,6 +234,20 @@ a floor can be trusted against, and today it is not:
   compares the plugin's manifest version against the registry's
   `latest_version` and nothing else.
 
+  *Fixed, in two parts.* `install_plugin` gained the gate in #431/#433, which
+  covers every registry-managed install and, through `_reinstall_with_rollback`,
+  the update path that re-downloads.
+  `update_plugin`'s git branch pulls in place and re-downloads nothing, so it
+  stayed ungated until `_gate_pulled_commit` closed it — checked after the pull
+  (the registry carries no floor field, so the incoming floor is unknowable
+  before it) and undone with `git reset --hard` to the pre-pull commit. That
+  route is rare in practice, since monorepo plugins install as archives; it was
+  closed because the sunset rule in the plugins repo's
+  `08-shared-sports-code.md` states as **condition 3** that the core enforces
+  the floor "at install/update time", and B6 rests on that being true rather
+  than merely written down. `install_from_url` — sideloading a plugin from a
+  URL — is still ungated.
+
 So B4 is: tag and release 3.2.0; make the tag, the release, and `__version__`
 agree, and keep them agreeing; reconsider the `< 2.0.0` skip; migrate manifests
 from `ledmatrix_min` to `ledmatrix_min_version`; and add the install/update
@@ -262,13 +276,21 @@ default is the more restrictive.
 `compatible_versions`. No manifest still carries it, so there is nothing to
 migrate there.)
 
-### B5 — adoption is safe by construction
+### B5 — the *fallback* is safe by construction; the modern path is not
+
+The heading matters, because the unqualified version of this claim is false and
+this document proves it two sections down: four of the eight adopted plugins
+shipped with scroll mode broken on a 3.2.0 core. What is safe by construction is
+narrower than "adoption".
 
 A plugin adopting core imports keeps its bundled copy and reaches it through the
-guarded import (see the Upgradability table above). On a core that ships the
+guarded import (see the Upgradability table above). On a core that doesn't ship
-module the plugin uses core code; on one that doesn't it falls back and behaves
+the module the plugin falls back and behaves exactly as it does today. That
-exactly as it does today. There is no version of this step that breaks a user,
+fallback compatibility — and only that — is safe by construction. On a core that
-which is why it does not wait for B6's gate.
+*does* ship the module, correctness is not automatic — object-level and scroll-mode validation
+(building both classes and comparing, per the retrospective below) is required
+to prove full behavior. There is no version of this step that breaks a user *on
+an old core*, which is why it does not wait for B6's gate.
 
 The hockey scroll-display pilot is **already validated**: adopted against a core
 carrying 3.2.0, `scroll_display.py` went from 691 to 289 lines and all 16 harness
@@ -297,7 +319,8 @@ been in users' hands long enough that the population running a core without it
 is small.** The bundled copies cost disk space; deleting them early costs
 scoreboards, silently. That trade is not close.
 
-Before the first sunset, add a **compatibility regression test**. It has to
+Before the first sunset, add a **compatibility regression test**. **Built:**
+core `test/test_sports_sunset_matrix.py` (#505). It has to
 cover four cases, not one — B5's safety claim and B6's failure mode are
 different propositions and only the second is obvious:
 
@@ -321,35 +344,144 @@ gate rather than trusting the failure to be noticed.
 The same suite should exercise the install/update gate, since it is the other
 half of the guarantee.
 
+### B6 — what actually happened
+
+**Ran 2026-09-01, across all eight scoreboards.** Held from 2026-08-05 to
+2026-09-01 on the argument below, which is kept because the reasoning applies to
+the next module, not because it is still in force.
+
+**The hold, and why it lifted.** The stated gate was evidence of 3.2.0 uptake —
+"a few months of it being the default download, or store-side install data".
+That evidence never arrived and could not: the core updates by
+`git pull --rebase`, so release-asset counts cannot measure uptake, and no
+store-side telemetry exists. What changed instead is that the *risk* the gate
+protected against was closed directly. The store now refuses a plugin whose
+floor exceeds the running core on **all three** routes in:
+
+| route | gated by |
+|---|---|
+| `install_plugin` — every path that re-downloads, `_reinstall_with_rollback` included | #431, #433 |
+| `update_plugin`'s git branch — pulls in place, re-downloads nothing | #508 |
+| `install_from_url` — sideloading | #510 |
+
+With all three closed a pre-3.2.0 user cannot receive a sunset plugin at all;
+they keep the version they already run. The population the hold existed to
+protect is protected by refusal rather than by a bundled copy — which is what
+the copy was standing in for.
+
+**What shipped.** Eight plugins, ~5,800 lines of frozen fallback deleted. Each:
+copy removed, guarded import collapsed to a plain one, floor raised to 3.2.0,
+`test_core_fallback.py` rewritten as `test_core_scroll.py` asserting the sunset
+rather than the fallback. `scripts/check_scroll_adoption.py` gained
+`sunset_violations` and a `SUNSET_PLUGINS` set naming all eight, so a
+resurrected copy or a returned guard fails CI.
+
+Delivered as plugins #346 (hockey, later folded into #351), #349 (football),
+#350 (baseball), #351 (the remaining six).
+
+**Two things found by doing it, both worth carrying forward:**
+
+- **Only one fallback held orchestration logic the core lacked.** baseball's
+  `_configure_scroll_helper` reinterpreted `scroll_speed` as pixels-per-*frame*
+  when `speed × delay` fell outside the 0.1–5.0 window — measured, 10–20×
+  faster than configured for a speed between 1.0 and 5.0. Standardised onto the
+  core's behaviour (honour the documented px/sec, clamp) rather than preserved.
+  Every other difference across the eight was a docstring, an unreachable
+  `scroll_helper is None` guard, or an equivalent diagnostic.
+- **Two tests had been leaning on the guard without anyone knowing.**
+  `soccer/test_live_screens.py` stubbed `src` in a way that shadowed the core,
+  so its guarded import fell back and the test had been exercising the *frozen
+  copy* rather than the shipping class since B5. Before sunsetting anything else
+  that carries a guarded core import, grep for tests that stub `src`.
+
+**The floor-raising traps still apply** to any future sunset: four plugins
+declare their floor top-level, where editing `versions[0]` is a silent no-op,
+and the floor has three live spellings (`min_ledmatrix_version`,
+`requires.min_ledmatrix_version`, `versions[].ledmatrix_min_version`, plus
+deprecated `ledmatrix_min`). See
+`src/plugin_system/compatibility.py:declared_min_version` for the resolution
+order any floor-raising tool must reproduce — and note the name is **inverted**
+between the top level and `versions[]`.
+
+**Still not adopted, deliberately:** `data_sources.py`, `game_renderer.py` and
+`base_odds_manager.py`. The standing decision held them until B6 closed; it now
+has, so they can be reconsidered — with B5's lesson applied, which is to build
+the object and diff rendered output rather than trust a static check.
+
+### B5 retrospective — what the adoption actually cost
+
+Recorded because it is the evidence behind the two decisions above, and because
+"the adoption went fine" is not what happened.
+
+**Four of the eight shipped with scroll mode broken** on a 3.2.0 core, and were
+repaired in plugins-repo #251. The restructure lifted the content methods
+verbatim but left the state they read off `self` behind: separator-icon
+constants (hockey, basketball, lacrosse) and the game-renderer cache (afl).
+hockey/basketball/lacrosse could not construct the scroll display at all; afl
+raised inside `prepare_scroll_content`, which the core base *catches*, so its
+only symptom was scroll mode silently drawing nothing.
+
+Three things are worth carrying forward:
+
+- **The bundled fallback did not protect anyone from this.** The break was on
+  the modern path, which the fallback never touches. Carrying the second copy
+  bought nothing against the actual defect while creating the divergence that
+  produced it. That is an argument *for* B6, not against it.
+- **Every gate was green.** The safety harness renders the scoreboard screens,
+  not scroll mode; `test_core_fallback.py` checked that methods existed and that
+  their *globals* resolved, and `self.NHL_SEPARATOR_ICON` is an attribute read,
+  invisible to an AST scan for `Name` loads. The fix was to stop reasoning about
+  source and **build the object**: construct both classes on both paths, compare
+  the separator icons they end up with, and assert the adopted class ends up
+  with every instance attribute the bundled one sets.
+- **Test what the change touches, not what is convenient to render.** Scroll
+  mode had no coverage because the harness could not reach it. A comparison
+  harness that renders the same games through both paths and diffs the pixels
+  needs no per-sport knowledge of the right answer, only that adopting core code
+  did not change it.
+
+**The ledger.** Before adoption, eight duplicated copies totalled 5,685 lines.
+After adoption plus the frozen legacy copies it was 10,610; removing the dead
+inline duplication (plugins #252) brought it to roughly 8,620. B6 would take it
+to about 3,300 including the shared core module — some 2,400 fewer than before
+this project started. **Until B6 runs, the adoption is net negative on disk**,
+and its one delivered user-visible gain is that adopted plugins honour the
+global `target_fps` instead of hardcoding ~100 FPS.
+
+### Decision: stop adopting further modules until B6 closes
+
+`data_sources.py` (9 copies), `game_renderer.py` (8) and `base_odds_manager.py`
+are the obvious next candidates. **Do not adopt them yet.** Each adoption adds
+carrying cost — a second copy to keep in step — against a payoff that is
+contingent on B6, and B6 is gated on an installed base we cannot currently
+measure. Consolidate what is already committed; revisit when B6 does.
+
 ## What's next
 
-In order. Each step is independently useful and independently revertible.
+Steps 1–5 of the original plan are **done**: 3.2.0 is tagged and published with
+a version number CI now asserts (#428), the compatibility gate is in
+`install_plugin` and reads `compatible_versions` as well as the floor
+(#431, #433), the newest manifest entry is required to use `ledmatrix_min_version`
+(plugins #244), and all eight plugins have adopted the scroll orchestration
+(plugins #245–#249, repaired in #251, tidied in #252).
 
-1. **Tag and publish v3.2.0.** The code is already on `main` (`21825cbf`).
+What actually remains, smallest first:
-   Nothing else blocks this, and it is what makes `ledmatrix_min_version:
+
-   "3.2.0"` refer to something real.
+1. **Soak the adoptions on hardware.** football and hockey have been run on a
-2. **Make the version number honest.** Have the release process assert that the
+   live rig through real games; baseball was watched through one earlier. The
-   tag, the GitHub release, and `src.__version__` agree — a check in CI is
+   rest are proven by harness, unit tests and pixel comparison. Out-of-season
-   cheaper than the confusion of the last two releases. Then revisit the
+   sports cannot be soaked until their season starts. When you do, **check the
-   `< 2.0.0` skip in `_warn_if_incompatible`, which currently silences the
+   rig's `*_display_mode` first** — a board in `switch` mode will happily load a
-   warning for the users who most need it.
+   sunset plugin and tell you nothing about the scroll code the sunset changed.
-3. **Add the compatibility gate** to `StoreManager.install_plugin` and
+2. **Cut 3.3.0.** Not required by B6 — its floors are 3.2.0, which is released —
-   `.update_plugin`: refuse a plugin whose declared floor exceeds
+   but `calendar` 1.2.3 floors at 3.3.0 for the device-authorization endpoints
-   `src.__version__`, and surface the reason in the store UI rather than only
+   that landed after 3.2.0, so it is un-installable until the release exists.
-   the log. This is the single change that turns the floor from documentation
+3. **Reconsider the held modules** (`data_sources.py`, `game_renderer.py`,
-   into a guarantee, and B6 depends on it.
+   `base_odds_manager.py`) now that the sunset has closed. `game_renderer.py` is
-4. **Migrate the manifests** to `ledmatrix_min_version`, and reconcile them with
+   the largest single duplication left: ~11,500 lines across eight plugins, with
-   `compatible_versions` (see above — that field is the required, canonical one,
+   ~36,500 more in the eight `sports.py`. Note that core already ships
-   and the gate does not read it yet). Currently 28 plugins spell the floor both
+   `src/base_classes/sports/` (~143KB, promoted in B1/B2) that **no plugin
-   ways across their `versions[]` entries, 12 use only the old spelling, and 2
+   imports** — check whether it has drifted before treating it as the target.
-   only the new. Scope the sweep to the nine sports plugins if a 42-plugin
-   version-bump wave isn't worth it — but the `compatible_versions` half has to
-   cover every manifest the gate can refuse, or define explicit legacy handling,
-   before the gate is allowed to block anything.
-5. **Run B5 adoption** — hockey, soccer, football, then the remaining six.
-   Bundled copies stay. Byte-identical harness output per plugin, then a soak.
-6. **Only then plan B6**, with the compatibility regression test described above
-   in CI first.
 
 ## How to keep this project healthy
 

docs/SSH_UNAVAILABLE_AFTER_INSTALL.md

@@ -20,7 +20,22 @@ The installation script:
 - Installs and configures `dnsmasq` (DHCP server for AP mode)
 - These services can interfere with normal WiFi client mode
 
-### 3. Reboot After Installation
+### 3. The Board Ran Out of Memory
+
+On a 512MB or 1GB board, memory exhaustion stops `sshd` being able to fork a
+session process. The connection is accepted and then closed immediately, before
+any banner:
+
+```text
+kex_exchange_identification: Connection closed by remote host
+```
+
+The giveaway is that the board is otherwise healthy — ping is clean and the web
+UI still responds — but nothing that needs to start a new process works, and
+the panel is usually dark. Only a power cycle clears it. See
+[LOW_MEMORY_BOARDS.md](LOW_MEMORY_BOARDS.md).
+
+### 4. Reboot After Installation
 
 If the script reboots the Pi (which it recommends), network services may restart in a different state, potentially triggering AP mode.
 
@@ -190,11 +205,23 @@ The web interface allows you to:
 
 ## Summary
 
-**SSH becomes unavailable because**:
+**SSH becomes unavailable because** — two unrelated causes, and they need
+different responses:
+
+*AP mode (most common):*
 - WiFi monitor service enables AP mode when WiFi disconnects
 - AP mode switches WiFi from client to access point mode
 - Pi loses connection to your original network
 
+*Memory exhaustion (low-memory boards):*
+- The board runs out of memory, so `sshd` cannot fork a session process
+- The connection is accepted and closed before any banner
+- Ping still answers and the web UI still responds, so it looks healthy
+- The panel is usually dark and the service cannot restart
+- **Only a power cycle clears this** — there is no remote recovery, because
+  every remote route needs a new process
+- Prevention and tuning: [LOW_MEMORY_BOARDS.md](LOW_MEMORY_BOARDS.md)
+
 **To regain SSH**:
 1. Connect to **LEDMatrix-Setup** AP network (password: `ledmatrix123`)
 2. SSH to `192.168.4.1`

first_time_install.sh

@@ -1419,9 +1419,16 @@ $ACTUAL_USER ALL=(ALL) NOPASSWD: $BASH_PATH $PROJECT_ROOT_DIR/scripts/fix_perms/
 EOF
 if [ -n "$JOURNALCTL_PATH" ]; then
     cat >> /tmp/ledmatrix_web_sudoers << EOF
-$ACTUAL_USER ALL=(ALL) NOPASSWD: $JOURNALCTL_PATH -u ledmatrix.service *
+# NOEXEC, because these rules end in a wildcard and journalctl starts a pager
-$ACTUAL_USER ALL=(ALL) NOPASSWD: $JOURNALCTL_PATH -u ledmatrix *
+# when its output is a terminal. From that pager (less) a "!sh" is a root
-$ACTUAL_USER ALL=(ALL) NOPASSWD: $JOURNALCTL_PATH -t ledmatrix *
+# shell -- the standard journalctl escalation. The web interface always passes
+# --no-pager, so nothing here needs it, but the rule cannot require a flag that
+# sits in the middle of the command line. NOEXEC stops the command executing
+# another program at all, which closes the hole without depending on wildcard
+# matching subtleties.
+$ACTUAL_USER ALL=(ALL) NOPASSWD:NOEXEC: $JOURNALCTL_PATH -u ledmatrix.service *
+$ACTUAL_USER ALL=(ALL) NOPASSWD:NOEXEC: $JOURNALCTL_PATH -u ledmatrix *
+$ACTUAL_USER ALL=(ALL) NOPASSWD:NOEXEC: $JOURNALCTL_PATH -t ledmatrix *
 EOF
 fi
 
@@ -1688,6 +1695,97 @@ else
     echo "✗ $CMDLINE_FILE not found; skipping isolcpus optimization"
 fi
 
+# Enable the memory cgroup controller (idempotent).
+# The Pi firmware boots with cgroup_disable=memory, so systemd's MemoryMax= is
+# accepted and silently ignored — the display service then has no ceiling, and
+# a runaway takes the whole board down (sshd can no longer fork, the panel goes
+# dark) rather than just restarting the one service.
+if [ "$SKIP_PERF" != "1" ] && [ -f "$CMDLINE_FILE" ]; then
+    # Both parameters are required for the memory controller, and they can get
+    # separated -- an image, another tool or a half-applied earlier run can
+    # leave one without the other. Checking only cgroup_enable=memory would
+    # report success while MemoryMax= silently does nothing, so each is checked
+    # and appended independently.
+    cgroup_missing=""
+    for cgroup_param in cgroup_enable=memory cgroup_memory=1; do
+        if ! grep -qw "$cgroup_param" "$CMDLINE_FILE"; then
+            cgroup_missing="$cgroup_missing $cgroup_param"
+        fi
+    done
+    if [ -z "$cgroup_missing" ]; then
+        echo "cgroup memory parameters already present in $CMDLINE_FILE"
+    else
+        echo "Adding${cgroup_missing} to $CMDLINE_FILE..."
+        cp "$CMDLINE_FILE" "$CMDLINE_FILE.bak" 2>/dev/null || true
+        # The kernel command line must stay on one line.
+        sed -i "1 s|\$|${cgroup_missing}|" "$CMDLINE_FILE"
+        echo "  Takes effect after reboot. Verify with:"
+        echo "    grep memory /sys/fs/cgroup/cgroup.controllers"
+    fi
+fi
+
+# Persist the journal (idempotent).
+# These images default to volatile storage: journald keeps everything in /run
+# (tmpfs), so every reboot destroys the logs — including the ones that would
+# explain why the board rebooted. Capped so an SD card is not worn out by logs.
+# A non-empty /var/log/journal does not prove journald is configured the way
+# this needs: the directory survives a switch back to volatile storage, and it
+# says nothing about whether a size cap is set. Read the effective
+# configuration instead, and only write the keys the user has not set
+# themselves so an explicit local limit is preserved.
+journald_effective() {
+    # systemd-analyze merges journald.conf with every drop-in; grep is the
+    # fallback for images that ship without it.
+    if command -v systemd-analyze >/dev/null 2>&1 &&
+       systemd-analyze cat-config systemd/journald.conf >/dev/null 2>&1; then
+        systemd-analyze cat-config systemd/journald.conf 2>/dev/null
+    else
+        cat /etc/systemd/journald.conf /etc/systemd/journald.conf.d/*.conf 2>/dev/null || true
+    fi
+}
+journald_conf="$(journald_effective)"
+# Both settings are optional, and stock images ship them commented out. grep
+# exits 1 on no match, which pipefail turns fatal under set -e — an absent
+# setting must read as empty, not abort the install.
+journald_storage="$(printf '%s\n' "$journald_conf" | grep -E '^[[:space:]]*Storage=' | tail -n1 | cut -d= -f2 | tr -d '[:space:]' || true)"
+journald_cap="$(printf '%s\n' "$journald_conf" | grep -E '^[[:space:]]*SystemMaxUse=' | tail -n1 | cut -d= -f2 | tr -d '[:space:]' || true)"
+
+if [ "$journald_storage" = "persistent" ] && [ -n "$journald_cap" ]; then
+    echo "Persistent journald storage already configured (SystemMaxUse=$journald_cap)"
+else
+    echo "Enabling persistent journald storage..."
+    mkdir -p /etc/systemd/journald.conf.d
+    {
+        echo "# Installed by LEDMatrix first_time_install.sh"
+        echo "[Journal]"
+        echo "Storage=persistent"
+        if [ -n "$journald_cap" ]; then
+            echo "# SystemMaxUse left to your existing setting ($journald_cap)"
+        else
+            # Capped so logs cannot wear out or fill an SD card.
+            echo "SystemMaxUse=64M"
+        fi
+    } > /etc/systemd/journald.conf.d/ledmatrix-persistent.conf
+    mkdir -p /var/log/journal
+    systemd-tmpfiles --create --prefix /var/log/journal >/dev/null 2>&1 || true
+    systemctl restart systemd-journald >/dev/null 2>&1 || true
+
+    # Drop-ins are applied in lexical order, so a locally added file that sorts
+    # after ledmatrix-persistent.conf (zz-local.conf and friends) still wins.
+    # Writing the file is not evidence it took effect -- re-read and say so
+    # plainly rather than reporting success we cannot confirm.
+    journald_now="$(journald_effective | grep -E '^[[:space:]]*Storage=' | tail -n1 | cut -d= -f2 | tr -d '[:space:]' || true)"
+    if [ "$journald_now" = "persistent" ]; then
+        echo "  Persistent journald storage active"
+    else
+        echo "  WARNING: journald storage is still '${journald_now:-unset}' after"
+        echo "  writing /etc/systemd/journald.conf.d/ledmatrix-persistent.conf."
+        echo "  Another drop-in that sorts later is overriding it. Check:"
+        echo "    systemd-analyze cat-config systemd/journald.conf | grep -n Storage="
+        echo "  Logs will not survive a reboot until that is resolved."
+    fi
+fi
+
 # Ensure dtparam=audio=off in config.txt (idempotent)
 if [ "$SKIP_PERF" = "1" ]; then
     : # skipped

scripts/check_plugin.py

@@ -41,6 +41,7 @@ from src.plugin_system.testing.loading import (  # noqa: E402
 )
 from src.plugin_system.testing.harness import (  # noqa: E402
     RenderResult, render_plugin_matrix, compare_to_goldens, write_goldens,
+    check_empty_claimed,
     check_scale_up,
 )
 from src.plugin_system.testing.sizes import (  # noqa: E402
@@ -115,6 +116,11 @@ def check_one(plugin_id: str, search_dirs: List[str], sizes, mock_data: Dict,
     declared = load_manifest(plugin_dir).get("display", {}).get("design_size", {})
     design_size = (int(declared.get("width", 128)), int(declared.get("height", 32)))
     fill_strict = spec.get("fill_check") == "strict"
+    # A mode that renders nothing without returning False is never skipped by
+    # the display controller, so it holds a blank panel for its whole duration.
+    # Warn-only by default: a scroll mode's first frame is legitimately its
+    # blank scroll-in buffer.
+    empty_strict = spec.get("empty_check") == "strict"
 
     # Every run: the base config, plus one per harness.json "variant" —
     # a config overlay with its own golden dir (e.g. adaptive layout mode
@@ -142,6 +148,7 @@ def check_one(plugin_id: str, search_dirs: List[str], sizes, mock_data: Dict,
             compare_to_goldens(results, golden_dir)
 
         check_scale_up(results, design_size=design_size, strict=fill_strict)
+        check_empty_claimed(results, strict=empty_strict)
 
         # Tag variant runs so the report and PNG dumps stay distinguishable.
         if variant_name:
@@ -178,6 +185,9 @@ def print_report(all_results: Dict[str, List[RenderResult]]) -> bool:
                     # warn-only underfill: big panel left mostly empty
                     ex, ey = r.fill_extent
                     detail += f" (fill warn: extent {ex:.0%}x{ey:.0%})"
+                if r.empty_claimed and r.empty_ok is None:
+                    detail += (f" (empty warn: drew nothing but display() returned"
+                               f" {r.display_returned!r}, so the mode is not skipped)")
             else:
                 everything_ok = False
                 if r.error is not None:
@@ -191,6 +201,11 @@ def print_report(all_results: Dict[str, List[RenderResult]]) -> bool:
                     ex, ey = r.fill_extent or (0.0, 0.0)
                     status = "FAIL"
                     detail = f" fill: extent {ex:.0%}x{ey:.0%} below required coverage"
+                elif r.empty_ok is False:
+                    status = "FAIL"
+                    detail = (f" drew nothing but display() returned"
+                              f" {r.display_returned!r}; return False so the"
+                              f" controller skips the mode")
                 else:
                     status, detail = "FAIL", ""
             print(f"  [{status}] {r.size_label:>7}  {r.mode}{detail}")

scripts/download_pixlet.sh

@@ -24,9 +24,29 @@ echo "========================================"
 # Auto-detect latest version if needed
 if [ "$PIXLET_VERSION" = "latest" ]; then
     echo "Detecting latest version..."
-    PIXLET_VERSION=$(curl -s "https://api.github.com/repos/${REPO}/releases/latest" | grep '"tag_name"' | sed -E 's/.*"([^"]+)".*/\1/')
+    # When this response arrives on a single line -- as it did on the device
-    if [ -z "$PIXLET_VERSION" ]; then
+    # where Starlark apps were failing -- `grep '"tag_name"'` matches the whole
-        echo "Failed to detect latest version, using fallback"
+    # document and a greedy `sed 's/.*"([^"]+)".*/\1/'` captures the LAST
+    # quoted token in it rather than the tag. That resolved to
+    # "mentions_count", which built a download URL for a release that does not
+    # exist. (The API is pretty-printed by default, which is why the old
+    # command looks correct when you try it by hand -- but the formatting is
+    # not something to depend on.) Match the field itself and take the value
+    # after it, which is right for either shape.
+    PIXLET_VERSION=$(curl -fsSL "https://api.github.com/repos/${REPO}/releases/latest" \
+        | grep -o '"tag_name"[[:space:]]*:[[:space:]]*"[^"]*"' \
+        | head -n1 \
+        | sed -E 's/.*:[[:space:]]*"([^"]*)".*/\1/')
+
+    # A wrong-but-non-empty value is what made the old bug silent, so check the
+    # shape rather than just that something came back. Anchored at both ends: a
+    # partial match would accept "v0.53garbage" or "0.53" and build a URL for a
+    # release that cannot exist, which is the failure this check is here to
+    # stop. Every tronbyt/pixlet release to date is vX.Y.Z; the optional suffix
+    # leaves room for a future -rc.1 or +build tag.
+    if ! printf '%s' "$PIXLET_VERSION" \
+        | grep -qE '^v[0-9]+\.[0-9]+\.[0-9]+([-+][0-9A-Za-z.-]+)?$'; then
+        echo "Could not detect the latest version (got: '${PIXLET_VERSION:-<empty>}'), using fallback"
         PIXLET_VERSION="v0.50.2"
     fi
 fi
@@ -67,8 +87,26 @@ download_binary() {
     temp_dir=$(mktemp -d -p "$PROJECT_ROOT" -t pixlet_download.XXXXXXXXXX)
     local temp_file="$temp_dir/$archive_name"
 
-    if ! curl -L -o "$temp_file" "$url" 2>/dev/null; then
+    # -f so an HTTP error is a failure. Without it curl writes the 404 body
-        echo "✗ Failed to download $arch"
+    # to the file and exits 0, and the first sign of trouble is tar saying
+    # "not in gzip format" about what is actually a page of HTML.
+    if ! curl -fL -o "$temp_file" "$url" 2>/dev/null; then
+        echo "✗ Failed to download $arch from $url"
+        rm -rf "$temp_dir"
+        return 1
+    fi
+
+    # Belt and braces: a mirror or proxy can return 200 with an error page.
+    if ! gzip -t "$temp_file" 2>/dev/null; then
+        echo "✗ Downloaded file is not a gzip archive: $url"
+        # These bytes come from whatever answered the request, so strip
+        # everything non-printable before echoing them: an error page carrying
+        # terminal escapes would otherwise be able to rewrite this output or
+        # bury it in a CI log. Printable characters are kept rather than
+        # hex-encoding the lot, because "<!DOCTYPE html>" is the diagnostic.
+        local first_bytes
+        first_bytes=$(head -c 60 "$temp_file" | tr -cd '[:print:]')
+        printf '  (first bytes: %s)\n' "$first_bytes"
         rm -rf "$temp_dir"
         return 1
     fi

scripts/install/configure_web_sudo.sh

@@ -100,10 +100,15 @@ TEMP_SUDOERS="/tmp/ledmatrix_web_sudoers_$$"
     echo "$WEB_USER ALL=(ALL) NOPASSWD: $SYSTEMCTL_PATH restart ledmatrix-web.service"
 
     # Optional: journalctl (non-critical — skip if not found)
+    #
+    # NOEXEC, matching first_time_install.sh. These rules end in a wildcard and
+    # journalctl starts a pager, so without it the caller can reach a shell:
+    # less runs "!command" as the user the pager belongs to, which here is
+    # root. NOEXEC stops the granted command executing anything of its own.
     if [ -n "$JOURNALCTL_PATH" ]; then
-        echo "$WEB_USER ALL=(ALL) NOPASSWD: $JOURNALCTL_PATH -u ledmatrix.service *"
+        echo "$WEB_USER ALL=(ALL) NOPASSWD:NOEXEC: $JOURNALCTL_PATH -u ledmatrix.service *"
-        echo "$WEB_USER ALL=(ALL) NOPASSWD: $JOURNALCTL_PATH -u ledmatrix *"
+        echo "$WEB_USER ALL=(ALL) NOPASSWD:NOEXEC: $JOURNALCTL_PATH -u ledmatrix *"
-        echo "$WEB_USER ALL=(ALL) NOPASSWD: $JOURNALCTL_PATH -t ledmatrix *"
+        echo "$WEB_USER ALL=(ALL) NOPASSWD:NOEXEC: $JOURNALCTL_PATH -t ledmatrix *"
     fi
 
     # Required: python3, bash

scripts/install/configure_wifi_permissions.sh

@@ -25,9 +25,44 @@ if [ "$EUID" -eq 0 ]; then
     exit 1
 fi
 
+# Resolve command paths against a fixed PATH, and check what we resolved.
+#
+# Every path found here is written into a sudoers file as a NOPASSWD grant, so
+# whoever controls the binary at that path controls root. first_time_install.sh
+# re-execs itself with `sudo -E`, which preserves the invoking user's
+# environment -- PATH included -- so without pinning it, `which nmcli` can
+# resolve to anything on that PATH: a writable directory early in it turns a
+# compromise of the low-privilege web user into permanent root.
+PATH=/usr/sbin:/usr/bin:/sbin:/bin
+export PATH
+
+# A binary named in a sudoers rule must be root-owned and writable by nobody
+# else, or the grant hands root to whoever can rewrite it.
+require_trusted_binary() {
+    local label="$1" path="$2"
+    if [ ! -x "$path" ]; then
+        echo "✗ $label: $path is not an executable file"
+        exit 1
+    fi
+    local owner perms
+    owner=$(stat -c '%u' "$path") || exit 1
+    perms=$(stat -c '%a' "$path") || exit 1
+    if [ "$owner" != "0" ]; then
+        echo "✗ $label: $path is not owned by root (uid $owner); refusing to"
+        echo "  grant it NOPASSWD sudo."
+        exit 1
+    fi
+    # Group- or world-writable means someone other than root can replace it.
+    case "$perms" in
+        *[2367]) echo "✗ $label: $path is writable by group or other ($perms);"
+                 echo "  refusing to grant it NOPASSWD sudo."
+                 exit 1 ;;
+    esac
+}
+
 # Get the full paths to commands
-NMCLI_PATH=$(which nmcli || echo "/usr/bin/nmcli")
+NMCLI_PATH=$(command -v nmcli || echo "/usr/bin/nmcli")
-SYSTEMCTL_PATH=$(which systemctl)
+SYSTEMCTL_PATH=$(command -v systemctl)
 
 echo "Command paths:"
 echo "  nmcli: $NMCLI_PATH"
@@ -37,6 +72,18 @@ echo "  systemctl: $SYSTEMCTL_PATH"
 echo ""
 echo "Step 1: Configuring sudo permissions for nmcli..."
 SUDOERS_FILE="/etc/sudoers.d/ledmatrix_wifi"
+SYSCTL_PATH=$(command -v sysctl || echo /usr/sbin/sysctl)
+NFT_PATH=$(command -v nft || echo /usr/sbin/nft)
+RFKILL_PATH=$(command -v rfkill || echo /usr/sbin/rfkill)
+MKDIR_PATH=$(command -v mkdir || echo /usr/bin/mkdir)
+
+# Checked before any of them reaches the sudoers file.
+require_trusted_binary "nmcli"     "$NMCLI_PATH"
+require_trusted_binary "systemctl" "$SYSTEMCTL_PATH"
+require_trusted_binary "sysctl"    "$SYSCTL_PATH"
+require_trusted_binary "nft"       "$NFT_PATH"
+require_trusted_binary "rfkill"    "$RFKILL_PATH"
+require_trusted_binary "mkdir"     "$MKDIR_PATH"
 
 # Create a temporary sudoers file using mktemp (handles permissions better)
 TEMP_SUDOERS=$(mktemp) || {
@@ -62,6 +109,36 @@ $WEB_USER ALL=(ALL) NOPASSWD: $SYSTEMCTL_PATH start dnsmasq
 $WEB_USER ALL=(ALL) NOPASSWD: $SYSTEMCTL_PATH stop dnsmasq
 $WEB_USER ALL=(ALL) NOPASSWD: $SYSTEMCTL_PATH restart dnsmasq
 $WEB_USER ALL=(ALL) NOPASSWD: $SYSTEMCTL_PATH restart NetworkManager
+# The captive portal turns IP forwarding on while the access point is up and
+# restores the previous value when it comes down (wifi_manager._setup_iptables_
+# redirect / _teardown_iptables_redirect). Without this rule that sudo call
+# needs a password, so forwarding stays off and clients associate to the AP but
+# cannot route. It goes unnoticed on a stock Raspberry Pi image, where
+# /etc/sudoers.d/010_pi-nopasswd grants the default user blanket NOPASSWD and
+# masks every gap in this file -- it only bites once that blanket rule is
+# removed.
+$WEB_USER ALL=(ALL) NOPASSWD: $SYSCTL_PATH -w net.ipv4.ip_forward=0
+$WEB_USER ALL=(ALL) NOPASSWD: $SYSCTL_PATH -w net.ipv4.ip_forward=1
+# The portal's redirect lives in its own nftables table, created when the AP
+# comes up and deleted when it goes down, and the radio has to be unblocked
+# before the AP can start at all. Same story as the sysctl rules above: called
+# with sudo, never granted here, and invisible on a stock Pi image.
+$WEB_USER ALL=(ALL) NOPASSWD: $NFT_PATH add table ip ledmatrix
+$WEB_USER ALL=(ALL) NOPASSWD: $NFT_PATH delete table ip ledmatrix
+$WEB_USER ALL=(ALL) NOPASSWD: $RFKILL_PATH unblock wifi
+# NetworkManager's dnsmasq drop-in directory, exact path.
+$WEB_USER ALL=(ALL) NOPASSWD: $MKDIR_PATH -p /etc/NetworkManager/dnsmasq-shared.d
+#
+# iptables is deliberately NOT granted here. Its rules are built from the live
+# interface name and port, so a rule covering them needs a trailing wildcard --
+# and `iptables --modprobe=/path/to/anything` runs that path as root, so
+# `NOPASSWD: iptables *` is a root shell for the web user by another name. That
+# is a worse outcome than the gap it would close, which today is masked anyway
+# by the blanket NOPASSWD rule on stock Pi images.
+#
+# Closing it safely means a wrapper script that builds the rules itself and
+# takes only an interface and a port, granted the way safe_plugin_rm.sh already
+# is. That belongs in its own change rather than being smuggled into this one.
 
 # Allow copying hostapd and dnsmasq config files into place
 $WEB_USER ALL=(ALL) NOPASSWD: /usr/bin/cp /tmp/hostapd.conf /etc/hostapd/hostapd.conf

scripts/render_plugin.py

@@ -52,6 +52,10 @@ def main() -> int:
     parser.add_argument('--height', type=int, default=32, help='Display height (default: 32)')
     parser.add_argument('--skip-update', action='store_true',
                         help='Skip calling update() (render display only)')
+    parser.add_argument('--display-mode', default=None,
+                        help='Display mode to render, for plugins that declare '
+                             'more than one in their manifest (e.g. nrl_live). '
+                             'Omitted, the plugin picks its own default.')
 
     args = parser.parse_args()
 
@@ -141,7 +145,25 @@ def main() -> int:
         except Exception as e:
             logger.warning("update() raised: %s — continuing to display()", e)
 
+    # A plugin that declares several display modes usually renders nothing
+    # useful without being told which one to draw: the scoreboards keep their
+    # state on per-mode sub-managers and their no-argument path returns False.
+    # Only pass the argument when asked for, so the many plugins whose display()
+    # takes no display_mode keep working untouched.
     try:
+        if args.display_mode:
+            try:
+                plugin_instance.display(display_mode=args.display_mode,
+                                        force_clear=True)
+            except TypeError as error:
+                if ("unexpected keyword argument" not in str(error)
+                        or "display_mode" not in str(error)):
+                    raise
+                logger.warning(
+                    "%s.display() does not accept display_mode; rendering its "
+                    "default screen instead", args.plugin)
+                plugin_instance.display(force_clear=True)
+        else:
             plugin_instance.display(force_clear=True)
         logger.debug("display() completed")
     except Exception as e:

src/__init__.py

@@ -4,5 +4,5 @@ LEDMatrix Display System
 Core source package for the LED Matrix Display project.
 """
 
-__version__ = "3.2.0"
+__version__ = "3.3.0"
 

src/background_data_service.py

@@ -14,11 +14,12 @@ Key Features:
 - Memory-efficient data storage
 """
 
+import itertools
 import time
 import logging
 import threading
 import requests
-from typing import Dict, Any, Optional, Callable
+from typing import Dict, Any, Optional, Callable, List
 from dataclasses import dataclass, field
 from enum import Enum
 import queue
@@ -50,14 +51,33 @@ class FetchRequest:
     max_retries: int = 3
     priority: int = 1  # Higher number = higher priority
     callback: Optional[Callable] = None
+    # Callbacks from submitters that JOINED this fetch instead of starting a
+    # duplicate one. The primary `callback` above belongs to whoever created
+    # the request; these belong to everyone who asked for the same cache_key
+    # while it was still in flight.
+    extra_callbacks: List[Callable] = field(default_factory=list)
     created_at: float = field(default_factory=time.time)
     status: FetchStatus = FetchStatus.PENDING
+    # Set once the worker has decided this response will be cached, while it
+    # still holds the lock. From that point cancelling is refused: the write
+    # is already authorised, and abandoning it here would put the payload in
+    # the cache with the callbacks suppressed -- joiners waiting forever for a
+    # fetch that did, in fact, succeed.
+    commit_claimed: bool = False
     result: Optional[Any] = None
     error: Optional[str] = None
 
 @dataclass
 class FetchResult:
-    """Result of a background fetch operation."""
+    """Result of a background fetch operation.
+
+    ``data`` survives on the stored result only for requests submitted without
+    a ``callback``, where polling ``get_result()`` is the sole way to collect
+    it. When a callback was given, the payload has already been delivered and
+    the service releases it -- see :meth:`BackgroundDataService._release_payload`.
+    Either way the data remains in the cache under the request's ``cache_key``,
+    which is where consumers read it from.
+    """
     request_id: str
     success: bool
     data: Optional[Any] = None
@@ -66,6 +86,10 @@ class FetchResult:
     fetch_time: float = 0.0
     retry_count: int = 0
     completed_at: float = field(default_factory=time.time)  # Timestamp when request completed
+    # The request's final status, recorded so a finished request can still be
+    # reported accurately. Without it a caller can only be told COMPLETED or
+    # FAILED, which turns "you cancelled this" into "this errored".
+    final_status: Optional[FetchStatus] = None
 
 class BackgroundDataService:
     """
@@ -90,6 +114,20 @@ class BackgroundDataService:
         
         # Thread management
         self.executor = ThreadPoolExecutor(max_workers=max_workers, thread_name_prefix="BackgroundData")
+        # cache_key -> request_id for fetches currently in flight. Submitting
+        # the same key twice used to start two identical fetches: request_id
+        # carries a millisecond timestamp, so every submit looked new, and
+        # active_requests is keyed by it rather than by what is being fetched.
+        # On a real board the season-schedule key is requested by both the
+        # Recent and the Upcoming manager, which miss the cache in the same
+        # millisecond and each download and parse the same payload.
+        self._inflight_by_cache_key: Dict[str, str] = {}
+        # request_id was sport_year_milliseconds, which is not unique: two
+        # submits inside the same millisecond produced the SAME id, so one
+        # silently replaced the other in active_requests and completed_requests.
+        # Rare before, but dedupe hands this id back to every joiner as their
+        # handle for get_result(), so it has to be unique. A counter is enough.
+        self._request_seq = itertools.count()
         self.active_requests: Dict[str, FetchRequest] = {}
         self.completed_requests: Dict[str, FetchResult] = {}
         self.request_queue = queue.PriorityQueue()
@@ -177,7 +215,9 @@ class BackgroundDataService:
         if cache_key is None:
             cache_key = self.get_sport_cache_key(sport)
         
-        request_id = f"{sport}_{year}_{int(time.time() * 1000)}"
+        with self._lock:
+            request_id = (f"{sport}_{year}_{int(time.time() * 1000)}"
+                          f"_{next(self._request_seq)}")
         
         # Check cache first
         cached_data = self.cache_manager.get(cache_key)
@@ -191,6 +231,10 @@ class BackgroundDataService:
                     cached=True,
                     fetch_time=0.0
                 )
+                # Filed before the callback runs, as it always was: a callback
+                # that queries get_result()/is_request_complete() for its own
+                # request must still find it. Releasing afterwards mutates the
+                # same object the dict holds.
                 self.completed_requests[request_id] = result
 
                 if callback:
@@ -198,6 +242,7 @@ class BackgroundDataService:
                         callback(result)
                     except Exception as e:
                         logger.error(f"Error in callback for request {request_id}: {e}")
+                    self._release_payload(result)
 
                 logger.debug(f"Cache hit for {sport} {year} data")
                 return request_id
@@ -218,7 +263,29 @@ class BackgroundDataService:
         )
         
         with self._lock:
+            existing_id = self._inflight_by_cache_key.get(cache_key)
+            existing = self.active_requests.get(existing_id) if existing_id else None
+            if existing_id and existing is None:
+                # Stranded index entry: the request it names is gone. Drop it and
+                # fetch normally. Looking the request up rather than trusting the
+                # id is what stops a stale entry wedging a key forever.
+                del self._inflight_by_cache_key[cache_key]
+            if existing is not None:
+                # Someone is already fetching this key. Ride along rather than
+                # duplicating the download, the parse and the resident copy.
+                if callback:
+                    existing.extra_callbacks.append(callback)
+                self.stats['deduplicated_requests'] = (
+                    self.stats.get('deduplicated_requests', 0) + 1
+                )
+                logger.info(
+                    "Joined in-flight fetch %s for %s (cache_key=%s) instead of "
+                    "starting a duplicate", existing_id, sport, cache_key
+                )
+                return existing_id
+
             self.active_requests[request_id] = request
+            self._inflight_by_cache_key[cache_key] = request_id
             self.stats['total_requests'] += 1
             self.stats['cache_misses'] += 1
         
@@ -243,7 +310,31 @@ class BackgroundDataService:
         
         try:
             with self._lock:
+                # A request cancelled while it sat in the executor queue must
+                # stay cancelled. Overwriting the status here undid the cancel
+                # outright: the worker went on to download, cache and call back
+                # for work the caller had already withdrawn.
+                if request.status == FetchStatus.CANCELLED:
+                    cancelled_before_start = True
+                else:
+                    cancelled_before_start = False
                     request.status = FetchStatus.IN_PROGRESS
+            if cancelled_before_start:
+                logger.info(
+                    "Request %s was cancelled before its worker started; "
+                    "skipping the fetch entirely", request.id
+                )
+                # Assign before returning: the finally block stores `result`
+                # in completed_requests, so building a fresh one here would
+                # file the untouched placeholder instead of this outcome.
+                result = FetchResult(
+                    request_id=request.id,
+                    success=False,
+                    error="cancelled",
+                    fetch_time=time.time() - start_time,
+                    retry_count=request.retry_count
+                )
+                return result
             
             logger.info(f"Starting background fetch for {request.sport} {request.year}")
             
@@ -269,6 +360,37 @@ class BackgroundDataService:
             # Log data validation
             logger.debug(f"Validated {len(events)} events for {request.sport} {request.year}")
             
+            # A cancelled request must not commit anything. Cancelling
+            # releases the cache_key, so a replacement fetch for the same key
+            # may already be in flight or finished -- writing this response to
+            # the cache now would overwrite fresher data with the response
+            # nobody wanted. The worker has no way to abort the HTTP call, so
+            # this is where the work gets discarded.
+            with self._lock:
+                cancelled = request.status == FetchStatus.CANCELLED
+                if not cancelled:
+                    # Claim the commit in the same critical section that read
+                    # the status, so a cancel cannot slip in between the check
+                    # and the cache write below. The write itself stays outside
+                    # the lock: it serialises a multi-megabyte payload to the
+                    # SD card, and holding the service lock across that would
+                    # stall every submit, status query and cancel behind it.
+                    request.commit_claimed = True
+            if cancelled:
+                logger.info(
+                    "Discarding response for cancelled request %s; %s may "
+                    "already belong to a replacement fetch",
+                    request.id, request.cache_key
+                )
+                result = FetchResult(
+                    request_id=request.id,
+                    success=False,
+                    error="cancelled",
+                    fetch_time=time.time() - start_time,
+                    retry_count=request.retry_count
+                )
+                return result
+
             # Cache the data
             self.cache_manager.set(request.cache_key, data)
             
@@ -294,6 +416,11 @@ class BackgroundDataService:
             logger.error(f"Failed to fetch {request.sport} {request.year} data: {error_msg}")
             
             with self._lock:
+                # Don't relabel a cancelled request. The callback gate in the
+                # finally block only suppresses CANCELLED, so promoting it to
+                # FAILED here delivered an error callback for a fetch nobody
+                # was waiting on any more.
+                if request.status != FetchStatus.CANCELLED:
                     request.status = FetchStatus.FAILED
                 request.error = error_msg
             
@@ -308,9 +435,26 @@ class BackgroundDataService:
         finally:
             # Store result and clean up
             with self._lock:
+                result.final_status = request.status
                 self.completed_requests[request.id] = result
                 if request.id in self.active_requests:
                     del self.active_requests[request.id]
+                # Stop accepting joiners and take the callback list in the same
+                # critical section. A submitter that arrives after this point
+                # finds no in-flight entry and either hits the cache (written
+                # above, before the result was built) or starts a fresh fetch --
+                # what it must never do is join a fetch whose callbacks have
+                # already run and then never be called.
+                if self._inflight_by_cache_key.get(request.cache_key) == request.id:
+                    del self._inflight_by_cache_key[request.cache_key]
+                # A cancelled request delivers nothing: its joiners were told
+                # about a fetch that has been abandoned, and a replacement will
+                # call them via its own request.
+                if request.status == FetchStatus.CANCELLED:
+                    callbacks = []
+                else:
+                    callbacks = ([request.callback] if request.callback else [])
+                    callbacks.extend(request.extra_callbacks)
                 
                 # Update statistics
                 if result.success:
@@ -327,15 +471,54 @@ class BackgroundDataService:
             # Periodic cleanup after storing result
             self._cleanup_completed_requests()
             
-            # Call callback if provided
+            # Call every callback: the original submitter's and any that joined
-            if request.callback:
+            # this fetch. One raising must not stop the others being delivered.
+            for cb in callbacks:
                 try:
-                    request.callback(result)
+                    cb(result)
                 except Exception as e:
                     logger.error(f"Error in callback for request {request.id}: {e}")
 
+            # Released AFTER the loop, not inside it. Every callback here holds
+            # the same FetchResult, so releasing per-delivery handed the first
+            # one the data and every joiner `result.data is None` -- which is
+            # not a quiet degradation: they read `result.data.get('events')` and
+            # raise AttributeError, which this very loop catches and logs, so
+            # the symptom was one ERROR line and a manager that silently never
+            # got its schedule. Deduplication is the normal case, not a corner:
+            # a sport's recent, upcoming and live managers all ride one season
+            # fetch.
+            #
+            # Guarded on `callbacks`, because a request submitted without one
+            # has no other way to collect its payload than polling get_result().
+            # The old per-delivery release got that right by accident: an empty
+            # list never entered the loop body.
+            if callbacks:
+                self._release_payload(result)
+                request.result = None
+
         return result
 
+    @staticmethod
+    def _release_payload(result: FetchResult) -> None:
+        """Drop a delivered payload, keeping the result's status and timings.
+
+        Only called once EVERY callback has been handed the data -- callers
+        that joined an in-flight fetch share this object, so releasing between
+        deliveries strips the payload out from under the ones still queued.
+        Consumers read fetched data back from the cache under ``cache_key``;
+        the copy carried
+        here was pinning a parsed season schedule -- 946 games for NCAA
+        football, roughly a tenth of total RAM on a 1GB Pi -- in memory until
+        the hourly sweep.
+
+        The cache-hit path matters most: it runs once per update interval per
+        sport, mints a fresh request_id each time, and a memory-tier miss
+        re-parses the payload from disk. Those were genuinely separate copies
+        accumulating toward the 500-entry cap, not shared references.
+        """
+        result.data = None
+    
     def _make_request_with_retry(self, request: FetchRequest) -> requests.Response:
         """
         Make HTTP request with retry logic and exponential backoff.
@@ -422,6 +605,8 @@ class BackgroundDataService:
                 return self.active_requests[request_id].status
             elif request_id in self.completed_requests:
                 result = self.completed_requests[request_id]
+                if result.final_status is not None:
+                    return result.final_status
                 return FetchStatus.COMPLETED if result.success else FetchStatus.FAILED
             return None
     
@@ -438,8 +623,22 @@ class BackgroundDataService:
         with self._lock:
             if request_id in self.active_requests:
                 request = self.active_requests[request_id]
+                if request.commit_claimed:
+                    # Too late: the worker holds an authorised commit. Report
+                    # the failure rather than half-cancelling a request whose
+                    # data is about to land in the cache.
+                    logger.debug(
+                        "Not cancelling %s: its response is already being "
+                        "committed", request_id
+                    )
+                    return False
                 request.status = FetchStatus.CANCELLED
                 del self.active_requests[request_id]
+                # Cancelling is the other way a request leaves active_requests,
+                # so the in-flight index has to be released here too or the key
+                # stays pointed at a request that no longer exists.
+                if self._inflight_by_cache_key.get(request.cache_key) == request_id:
+                    del self._inflight_by_cache_key[request.cache_key]
                 logger.info(f"Cancelled request {request_id}")
                 return True
             return False

src/base_classes/sports/modes.py

@@ -145,8 +145,7 @@ class SportsUpcoming(SportsCore):
                     if (game['home_abbr'] in self.favorite_teams or 
                         game['away_abbr'] in self.favorite_teams):
                         favorite_games_found += 1
-                    if self.show_odds:
+                    # Odds are NOT fetched here -- see after selection below.
-                        self._fetch_odds(game)
 
             # Enhanced logging for debugging
             self.logger.info(f"Found {all_upcoming_games} total upcoming games in data")
@@ -190,6 +189,20 @@ class SportsUpcoming(SportsCore):
                 # Limit to the specified number of upcoming games
                 team_games = team_games[:self.upcoming_games_to_show]
 
+            # Odds are fetched here, for the games that survived selection,
+            # rather than in the loop that collects them. That loop walks every
+            # upcoming game in the schedule window, and for a college league
+            # the window is enormous -- a live rig logged 946 upcoming games in
+            # one cycle and displayed 1 of them. The comment up there claimed
+            # odds were fetched "only for games that will be displayed", but
+            # the only narrowing it applied was show_favorite_teams_only, which
+            # is not the default; in the usual case nothing narrowed it at all
+            # and every game cost a separate ESPN request on a Pi that is also
+            # driving the panel.
+            if self.show_odds:
+                for game in team_games:
+                    self._fetch_odds(game)
+
             # Log changes or periodically
             should_log = (
                  current_time - self.last_log_time >= self.log_interval or

src/base_odds_manager.py

@@ -12,6 +12,8 @@ Follows LEDMatrix configuration management patterns:
 """
 
 import logging
+import time
+
 import requests
 import json
 from typing import Dict, Any, Optional, List
@@ -43,9 +45,34 @@ class BaseOddsManager:
         self.logger = logging.getLogger(__name__)
         self.base_url = "https://sports.core.api.espn.com/v2/sports"
 
+        # This path used a bare requests.get, so it identified itself as
+        # python-requests/x.y -- the one thing ESPN is known to reject. Around
+        # 2026-08-04 it began 403ing browser strings and bare custom tokens
+        # alike; what it accepts is a token with a URL that says who is
+        # calling. Every other ESPN caller in the tree already sends this
+        # (src/common/api_helper.py, src/base_classes/data_sources.py); the
+        # odds path was simply missed, and it is the one whose failures cost
+        # the caller its whole update budget.
+        #
+        # Deliberately no retry adapter, unlike api_helper: retries multiply
+        # request_timeout, which is set to 5s precisely to stay inside that
+        # budget. One try, then the cooldown below.
+        self.session = requests.Session()
+        self.session.headers.update({
+            'User-Agent': 'LEDMatrix/1.0 (+https://github.com/ChuckBuilds/LEDMatrix)',
+            'Accept': 'application/json',
+        })
+        
         # Configuration with defaults
         self.update_interval = 3600  # 1 hour default
-        self.request_timeout = 30    # 30 seconds default
+        # Well under the plugin executor's 30s operation budget. At 30s a
+        # single stalled ESPN request consumed the entire budget and the whole
+        # update() was killed -- and odds are fetched per live game, inside the
+        # live update loop, with show_odds defaulting on. Losing one game's
+        # odds beats losing the update that carries every game's score.
+        self.request_timeout = 5
+        # Set when a request fails; until then, skip the network entirely.
+        self._skip_network_until = 0.0
         self.cache_ttl = 1800       # 30 minutes default
         
         # Load configuration if available
@@ -73,6 +100,14 @@ class BaseOddsManager:
         except Exception as e:
             self.logger.warning(f"Failed to load BaseOddsManager configuration: {e}")
     
+    # After a network failure, stop trying for this long and serve cache only.
+    # A short per-request timeout bounds one stall, but a full Sunday slate is
+    # ~16 games fetched in a loop, so 16 consecutive timeouts still blow the
+    # budget. When ESPN is unreachable it is unreachable for all of them, so
+    # the first failure is enough to know: skip the rest of this pass and try
+    # again shortly.
+    _FAILURE_COOLDOWN = 60.0
+
     def get_odds(self, sport: str | None, league: str | None, event_id: str,
                  update_interval_seconds: int = None) -> Optional[Dict[str, Any]]:
         """
@@ -101,6 +136,16 @@ class BaseOddsManager:
             self.logger.info(f"Using cached odds from ESPN for {cache_key}")
             return cached_data
 
+        if time.monotonic() < self._skip_network_until:
+            # A recent request failed, so ESPN is very likely still unreachable.
+            # Returning now keeps the caller's update inside its time budget
+            # instead of paying the timeout again for every remaining game.
+            self.logger.debug(
+                "Skipping odds fetch for %s: a recent request failed, holding off "
+                "for another %.0fs", cache_key,
+                self._skip_network_until - time.monotonic())
+            return None
+
         self.logger.info(f"Cache miss - fetching fresh odds from ESPN for {cache_key}")
 
         try:
@@ -117,10 +162,12 @@ class BaseOddsManager:
             url = f"{self.base_url}/{sport}/leagues/{espn_league}/events/{event_id}/competitions/{event_id}/odds"
             self.logger.info(f"Requesting odds from URL: {url}")
             
-            response = requests.get(url, timeout=self.request_timeout)
+            response = self.session.get(url, timeout=self.request_timeout)
             response.raise_for_status()
             raw_data = response.json()
 
+            self._skip_network_until = 0.0   # reachable again
+
             self.logger.debug(f"Received raw odds data from ESPN: {json.dumps(raw_data, indent=2)}")
             
             odds_data = self._extract_espn_data(raw_data)
@@ -140,7 +187,11 @@ class BaseOddsManager:
             return odds_data
 
         except requests.exceptions.RequestException as e:
-            self.logger.error(f"Error fetching odds from ESPN API for {cache_key}: {e}")
+            self._skip_network_until = time.monotonic() + self._FAILURE_COOLDOWN
+            self.logger.error(
+                "Error fetching odds from ESPN API for %s: %s. Holding off on odds "
+                "for %.0fs so a slate of games does not pay this timeout each.",
+                cache_key, e, self._FAILURE_COOLDOWN)
         except json.JSONDecodeError:
             self.logger.error(f"Error decoding JSON response from ESPN API for {cache_key}.")
         
@@ -163,19 +214,25 @@ class BaseOddsManager:
             item = data["items"][0]
             self.logger.debug(f"First item keys: {list(item.keys())}")
             
-            # The ESPN API returns odds data directly in the item, not in a providers array
+            # The ESPN API returns odds data directly in the item, not in a
-            # Extract the odds data directly from the item
+            # providers array. ESPN sends explicit JSON nulls for absent
+            # sides ("homeTeamOdds": null), so every level uses `or {}` —
+            # .get's default only applies when the key is missing entirely.
+            home = item.get("homeTeamOdds") or {}
+            away = item.get("awayTeamOdds") or {}
             extracted_data = {
                 "details": item.get("details"),
                 "over_under": item.get("overUnder"),
                 "spread": item.get("spread"),
                 "home_team_odds": {
-                    "money_line": item.get("homeTeamOdds", {}).get("moneyLine"),
+                    "money_line": home.get("moneyLine"),
-                    "spread_odds": item.get("homeTeamOdds", {}).get("current", {}).get("pointSpread", {}).get("value")
+                    "spread_odds": ((home.get("current") or {})
+                                    .get("pointSpread") or {}).get("value")
                 },
                 "away_team_odds": {
-                    "money_line": item.get("awayTeamOdds", {}).get("moneyLine"),
+                    "money_line": away.get("moneyLine"),
-                    "spread_odds": item.get("awayTeamOdds", {}).get("current", {}).get("pointSpread", {}).get("value")
+                    "spread_odds": ((away.get("current") or {})
+                                    .get("pointSpread") or {}).get("value")
                 }
             }
             self.logger.debug(f"Returning extracted odds data: {json.dumps(extracted_data, indent=2)}")
@@ -260,7 +317,11 @@ class BaseOddsManager:
         Returns:
             Formatted odds summary string
         """
-        if not self.is_odds_available(odds_data):
+        # Gate only on truly-empty / negative-cached data. is_odds_available
+        # deliberately ignores money lines (its callers decide whether to
+        # RENDER an odds widget), but a summary of money-line-only odds is
+        # still meaningful — the parts loop below handles them.
+        if not odds_data or odds_data.get('no_odds'):
             return "No odds available"
 
         parts = []

src/cache/disk_cache.py

@@ -14,6 +14,13 @@ import zlib
 from typing import Dict, Any, Optional, Protocol
 from datetime import datetime
 
+# How old an abandoned write's temp file must be before the sweep removes it.
+# A real write holds its temp file for milliseconds, so an hour is far beyond
+# any in-flight write while still clearing the same day's debris. Deliberately
+# not tied to the retention policies: those describe how long data stays
+# useful, and a half-written file was never useful.
+_ORPHAN_TEMP_MAX_AGE_SECONDS = 3600
+
 
 
 class CacheStrategyProtocol(Protocol):
@@ -112,6 +119,22 @@ class DiskCache:
                     record_ts = None
             
             now = time.time()
+
+            # An explicit per-entry ttl wins over the caller's max_age. The
+            # caller that wrote the record knows what its data is; max_age is
+            # inferred from substrings in the key ("live", "odds", "stock") and
+            # is only a fallback for records that never said. Until now the ttl
+            # was stored and ignored, so `set(key, data, ttl=...)` did nothing
+            # at all -- 48 plugin call sites and 4 in the core were writing a
+            # number no read path consulted.
+            effective_max_age = max_age
+            if isinstance(record, dict):
+                stored_ttl = record.get('ttl')
+                if isinstance(stored_ttl, (int, float)) and not isinstance(stored_ttl, bool) \
+                        and stored_ttl >= 0:
+                    effective_max_age = stored_ttl
+            max_age = effective_max_age
+
             # max_age=None means "never expires" (mirrors MemoryCache and the
             # cache_manager docstring). Guard it explicitly — otherwise the
             # comparison below raises TypeError and the record is treated as a
@@ -331,6 +354,23 @@ class DiskCache:
         """Get the cache directory path."""
         return self.cache_dir
     
+    @staticmethod
+    def _is_orphaned_temp(filename: str) -> bool:
+        """Whether a name is one of set()'s temp files rather than real data.
+
+        Matches only what this class creates: mkstemp with a prefix of
+        ".<cache filename>." , so ".weather.json.a1b2c3d4". The shape is
+        checked rather than just the leading dot, because this predicate
+        deletes things -- a stray dotfile someone left in the cache directory
+        is not ours to remove, and a completed ".json" never is either.
+        """
+        if not filename.startswith('.') or filename.endswith('.json'):
+            return False
+        head, sep, suffix = filename.rpartition('.json.')
+        # head is the key (non-empty after the leading dot), suffix is
+        # mkstemp's random component.
+        return bool(sep) and len(head) > 1 and bool(suffix)
+
     def cleanup_expired_files(self, cache_strategy: CacheStrategyProtocol, retention_policies: Dict[str, int]) -> Dict[str, Any]:
         """
         Clean up expired cache files based on retention policies.
@@ -365,12 +405,51 @@ class DiskCache:
             try:
                 with self._lock:
                     # Get snapshot of files while holding lock briefly
-                    filenames = [f for f in os.listdir(self.cache_dir) if f.endswith('.json')]
+                    entries = os.listdir(self.cache_dir)
             except OSError as list_error:
                 self.logger.error("Error listing cache directory %s: %s", self.cache_dir, list_error, exc_info=True)
                 stats['errors'] += 1
                 return stats
 
+            filenames = [f for f in entries if f.endswith('.json')]
+
+            # Sweep temp files abandoned by a write that never finished. set()
+            # removes its own in a finally, so these are the ones where the
+            # process died between mkstemp and os.replace -- a SIGKILL, a lost
+            # restart race, a power cut. Nothing ever collected them: they are
+            # named ".<key>.json.<random>", and the scan above only matches
+            # names ending in .json, so they accumulated indefinitely. Measured
+            # on a live rig: 76 files, 1,050 MB, 81% of the whole cache
+            # directory, the oldest six months old.
+            stats['orphan_temp_files_deleted'] = 0
+            for filename in (f for f in entries if self._is_orphaned_temp(f)):
+                # Counted as scanned like any other candidate, so files_deleted
+                # can never exceed files_scanned and the summary line reads
+                # honestly ("77/8864", not "77/0").
+                stats['files_scanned'] += 1
+                path = os.path.join(self.cache_dir, filename)
+                try:
+                    # An in-flight write lives for milliseconds, so anything
+                    # this old is certainly abandoned rather than in progress.
+                    if (current_time - os.path.getmtime(path)) <= _ORPHAN_TEMP_MAX_AGE_SECONDS:
+                        continue
+                    with self._lock:
+                        size = os.path.getsize(path)
+                        os.remove(path)
+                    stats['files_deleted'] += 1
+                    stats['orphan_temp_files_deleted'] += 1
+                    stats['space_freed_bytes'] += size
+                except FileNotFoundError:
+                    continue          # another sweep got there first
+                except OSError as e:
+                    stats['errors'] += 1
+                    self.logger.warning("Error deleting orphaned temp file %s: %s", filename, e)
+
+            if stats['orphan_temp_files_deleted']:
+                self.logger.info(
+                    "Removed %d abandoned cache temp file(s)",
+                    stats['orphan_temp_files_deleted'])
+            
             # Process files outside the lock to avoid blocking get/set operations
             for filename in filenames:
                 stats['files_scanned'] += 1

src/cache/memory_cache.py

@@ -4,11 +4,58 @@ Memory Cache
 Handles in-memory caching with TTL support, size limits, and automatic cleanup.
 """
 
+import os
 import time
 import threading
 import logging
 from typing import Dict, Any, Optional
 
+# Historical fixed ceiling, kept as the fallback when RAM cannot be read.
+DEFAULT_MAX_SIZE = 1000
+
+
+def _total_memory_mb() -> Optional[float]:
+    """Physical RAM in MB, or None where /proc/meminfo is unavailable."""
+    try:
+        with open('/proc/meminfo', 'r', encoding='utf-8') as fh:
+            for line in fh:
+                if line.startswith('MemTotal:'):
+                    return int(line.split()[1]) / 1024
+    except (OSError, ValueError, IndexError):
+        return None
+    return None
+
+
+def default_max_size() -> int:
+    """Entry ceiling scaled to this machine's RAM.
+
+    One fixed ceiling cannot serve both a 512 MB Pi Zero 2 W and an 8 GB Pi 5.
+    Entries here are parsed API payloads that routinely run tens of kilobytes
+    each, so a thousand of them is a comfortable cache on a large board and a
+    substantial fraction of total RAM on a small one — where the process
+    competing for that RAM is also driving the panel. Set
+    LEDMATRIX_CACHE_MAX_ENTRIES to override.
+    """
+    override = os.environ.get('LEDMATRIX_CACHE_MAX_ENTRIES')
+    if override:
+        try:
+            value = int(override)
+            if value > 0:
+                return value
+        except ValueError:
+            pass
+
+    total_mb = _total_memory_mb()
+    if total_mb is None:
+        return DEFAULT_MAX_SIZE
+    if total_mb < 1536:      # 512 MB and 1 GB boards
+        return 150
+    if total_mb < 3072:      # 2 GB
+        return 400
+    if total_mb < 6144:      # 4 GB
+        return 800
+    return 1500              # 8 GB and up
+
 
 class MemoryCache:
     """Manages in-memory cache with TTL and size limits."""
@@ -57,6 +104,16 @@ class MemoryCache:
             if timestamp is None:
                 return None
             
+            # An explicit per-entry ttl wins over the caller's max_age, matching
+            # DiskCache. max_age is inferred from substrings in the key and is
+            # only a fallback for records that did not say what they wanted.
+            record = self._cache[key]
+            if isinstance(record, dict):
+                stored_ttl = record.get('ttl')
+                if isinstance(stored_ttl, (int, float)) and not isinstance(stored_ttl, bool) \
+                        and stored_ttl >= 0:
+                    max_age = stored_ttl
+
             # Check expiration
             if max_age is not None and (now - timestamp) > max_age:
                 # Expired - remove it
@@ -77,6 +134,32 @@ class MemoryCache:
         with self._lock:
             self._cache[key] = value
             self._timestamps[key] = time.time()
+            # Enforce the ceiling here rather than leaving it to the periodic
+            # cleanup, which only runs every cleanup_interval seconds (300 by
+            # default). A burst of inserts between two sweeps could otherwise
+            # take the cache far past _max_size, which is the memory growth this
+            # limit exists to prevent -- and on a 1GB board that is the
+            # difference between a bounded cache and an unreachable Pi.
+            self._evict_over_limit_locked()
+
+    def _evict_over_limit_locked(self) -> int:
+        """Drop oldest entries until the cache is within _max_size.
+
+        Caller must hold self._lock. Returns the number of entries removed.
+        """
+        excess = len(self._cache) - self._max_size
+        if excess <= 0:
+            return 0
+        oldest = sorted(
+            self._timestamps.items(),
+            key=lambda item: float(item[1]) if isinstance(item[1], (int, float)) else 0.0
+        )
+        removed = 0
+        for key, _ in oldest[:excess]:
+            self._cache.pop(key, None)
+            self._timestamps.pop(key, None)
+            removed += 1
+        return removed
     
     def clear(self, key: Optional[str] = None) -> None:
         """
@@ -133,22 +216,8 @@ class MemoryCache:
                 self._timestamps.pop(key, None)
                 removed_count += 1
             
-            # Enforce size limit by removing oldest entries if cache is too large
+            # Same ceiling enforcement set() uses, so the two cannot drift.
-            if len(self._cache) > self._max_size:
+            removed_count += self._evict_over_limit_locked()
-                # Sort by timestamp (oldest first)
-                sorted_entries = sorted(
-                    self._timestamps.items(),
-                    key=lambda x: float(x[1]) if isinstance(x[1], (int, float)) else 0
-                )
-                
-                # Remove oldest entries until we're under the limit
-                excess_count = len(self._cache) - self._max_size
-                for i in range(excess_count):
-                    if i < len(sorted_entries):
-                        key = sorted_entries[i][0]
-                        self._cache.pop(key, None)
-                        self._timestamps.pop(key, None)
-                        removed_count += 1
             
             self._last_cleanup = current_time
             

src/cache_manager.py

@@ -33,7 +33,7 @@ import logging
 import threading
 import tempfile
 from src.exceptions import CacheError
-from src.cache.memory_cache import MemoryCache
+from src.cache.memory_cache import MemoryCache, default_max_size
 from src.cache.disk_cache import DiskCache
 from src.cache.cache_strategy import CacheStrategy
 from src.cache.cache_metrics import CacheMetrics
@@ -47,6 +47,20 @@ from src.cache.disk_cache import DateTimeEncoder  # noqa: F401 - deliberate re-e
 class CacheManager:
     """Manages caching of API responses to reduce API calls."""
 
+    # Which cache directories already have a cleanup thread in this process.
+    #
+    # The sweep is directory-scoped work -- it lists a directory and deletes
+    # from it -- so one per directory is the right number no matter how many
+    # managers exist. Nothing enforced that before: every instance started its
+    # own, and because the loop closes over `self`, a discarded manager could
+    # never be collected and its thread woke to re-scan the same directory
+    # every 24 hours for the life of the process. Startup validation runs
+    # twice and built a throwaway manager each time, so a display process
+    # carried three threads for one cache.
+    _cleanup_owners: Dict[str, 'CacheManager'] = {}
+    _cleanup_owners_lock = threading.Lock()
+
+
     def __init__(self) -> None:
         # Initialize logger first
         self.logger: logging.Logger = get_logger(__name__)
@@ -70,7 +84,9 @@ class CacheManager:
             self.logger.warning("ConfigManager not available, using default cache intervals")
         
         # Initialize cache components using composition
-        self._memory_cache_component = MemoryCache(max_size=1000, cleanup_interval=300.0)
+        self._memory_cache_component = MemoryCache(
+            max_size=default_max_size(), cleanup_interval=300.0
+        )
         self._disk_cache_component = DiskCache(cache_dir=self.cache_dir, logger=self.logger)
         self._strategy_component = CacheStrategy(config_manager=self.config_manager, logger=self.logger)
         self._metrics_component = CacheMetrics(logger=self.logger)
@@ -594,8 +610,10 @@ class CacheManager:
         Args:
             key: Cache key
             data: Data to cache
-            ttl: Optional time-to-live in seconds (stored for compatibility but
+            ttl: Time-to-live in seconds for this entry. Takes precedence over
-                 expiration is still controlled via max_age when reading)
+                 the max_age a reader would otherwise apply, which is inferred
+                 from the key and is only a fallback for entries that did not
+                 say. Omit it to keep that inferred behaviour.
         """
         cache_data = {
             'data': data,
@@ -716,11 +734,29 @@ class CacheManager:
             }
     
     def start_cleanup_thread(self) -> None:
-        """Start background thread for periodic disk cache cleanup."""
+        """Start background thread for periodic disk cache cleanup.
+
+        At most one thread per cache directory per process: the sweep is
+        directory-scoped, so a second one only duplicates the scan.
+        """
         if self._cleanup_thread and self._cleanup_thread.is_alive():
             self.logger.debug("Cleanup thread already running")
             return
 
+        with CacheManager._cleanup_owners_lock:
+            owner = CacheManager._cleanup_owners.get(self.cache_dir)
+            if owner is not None and owner is not self:
+                thread = owner._cleanup_thread
+                if thread is not None and thread.is_alive():
+                    self.logger.debug(
+                        "Cleanup thread for %s already owned by another cache "
+                        "manager in this process; not starting a second",
+                        self.cache_dir)
+                    return
+                # The owner's thread died or was stopped -- take over.
+            CacheManager._cleanup_owners[self.cache_dir] = self
+
+
         def cleanup_loop():
             """Background loop that runs cleanup periodically."""
             self.logger.info("Disk cache cleanup thread started (interval: %d hours)", 
@@ -768,6 +804,13 @@ class CacheManager:
         Signals the thread to stop and waits for it to finish (with timeout).
         This allows for clean shutdown during testing or application termination.
         """
+        # Release ownership first and unconditionally, so a manager that never
+        # started a thread (or whose thread already exited) cannot keep the
+        # directory claimed and block a live manager from sweeping it.
+        with CacheManager._cleanup_owners_lock:
+            if CacheManager._cleanup_owners.get(self.cache_dir) is self:
+                del CacheManager._cleanup_owners[self.cache_dir]
+
         if not self._cleanup_thread or not self._cleanup_thread.is_alive():
             self.logger.debug("Cleanup thread not running")
             return

src/common/api_helper.py

@@ -273,19 +273,34 @@ class APIHelper:
         """
         Clear cache data.
 
+        Uses CacheManager's real surface (clear_cache / delete /
+        list_cache_files); safely no-ops on managers without it. The old
+        implementation guarded on a nonexistent ``clear`` method, so it
+        silently never cleared anything.
+
         Args:
-            pattern: Optional pattern to match cache keys
+            pattern: Optional substring to match cache keys; only matching
+                entries are deleted.
         """
-        if self.cache_manager:
+        if not self.cache_manager:
-            if hasattr(self.cache_manager, 'clear'):
+            return
         if pattern:
-                    # Clear only keys matching pattern
+            if (hasattr(self.cache_manager, 'list_cache_files')
-                    keys = self.cache_manager.keys()
+                    and hasattr(self.cache_manager, 'delete')):
-                    for key in keys:
+                for entry in self.cache_manager.list_cache_files():
-                        if pattern in key:
+                    key = entry.get('key') if isinstance(entry, dict) else None
+                    if key and pattern in key:
                         self.cache_manager.delete(key)
             else:
+                self.logger.debug(
+                    "Cache manager lacks list_cache_files/delete; "
+                    "cannot clear by pattern")
+        elif hasattr(self.cache_manager, 'clear_cache'):
+            self.cache_manager.clear_cache()
+        elif hasattr(self.cache_manager, 'clear'):
             self.cache_manager.clear()
+        else:
+            self.logger.debug("Cache manager exposes no clear method; no-op")
     
     def _get_from_cache(self, key: str) -> Optional[Any]:
         """Get data from cache."""

src/common/config_helper.py

@@ -5,6 +5,7 @@ Handles configuration management and validation for LED matrix plugins.
 Extracted from LEDMatrix core to provide reusable functionality for plugins.
 """
 
+import copy
 import json
 import logging
 from pathlib import Path
@@ -160,17 +161,20 @@ class ConfigHelper:
             override_config: Configuration to merge in (takes precedence)
             
         Returns:
-            Merged configuration dictionary
+            Merged configuration dictionary (fully independent of both
+            inputs — a shallow copy would alias un-overridden nested dicts,
+            so mutating the result would mutate the caller's base config).
         """
-        merged = base_config.copy()
+        merged = copy.deepcopy(base_config)
 
         for key, value in override_config.items():
             if key in merged and isinstance(merged[key], dict) and isinstance(value, dict):
                 # Recursively merge nested dictionaries
                 merged[key] = self.merge_configs(merged[key], value)
             else:
-                # Override with new value
+                # Override with new value — deep-copied so mutating the
-                merged[key] = value
+                # merged result can't reach back into override_config.
+                merged[key] = copy.deepcopy(value)
         
         return merged
     

src/common/display_helper.py

@@ -115,18 +115,14 @@ class DisplayHelper:
         if home_logo and away_logo:
             self._draw_logos(main_img, home_logo, away_logo)
         
-        # Draw status/period text (top center)
+        # Draw one combined top line (period/status/clock all share y=1 —
-        if status_text or period_text:
+        # drawing them separately overprinted each other).
-            status_display = f"{period_text} {status_text}".strip()
+        top_line = " ".join(p for p in [period_text, status_text, clock] if p)
-            if status_display:
+        if top_line:
-                self._draw_centered_text(draw, status_display, 
+            self._draw_centered_text(draw, top_line,
                                    fonts.get('time', fonts.get('status')),
                                    y_position=1)
         
-        # Draw clock if available
-        if clock:
-            self._draw_centered_text(draw, clock, fonts.get('time'), y_position=1)
-        
         # Draw scores (center)
         score_text = f"{away_score}-{home_score}"
         self._draw_centered_text(draw, score_text, fonts.get('score'), 
@@ -153,12 +149,18 @@ class DisplayHelper:
         """
         Draw a ticker/scrolling text layout.
         
+        Renders a single static frame with the text at the left edge; the
+        caller advances the scroll by re-rendering or shifting. The
+        scroll_speed parameter is accepted for API compatibility but does
+        not affect this frame. (Previously the text was drawn starting at
+        x=display_width — entirely off-canvas — so every frame was blank.)
+
         Args:
             text: Text to display
             font: Font to use
             background_color: Background color
             text_color: Text color
-            scroll_speed: Pixels to scroll per frame
+            scroll_speed: Accepted for compatibility; unused per-frame
 
         Returns:
             PIL Image with ticker layout
@@ -166,11 +168,7 @@ class DisplayHelper:
         img = self.create_base_image(background_color)
         draw = ImageDraw.Draw(img)
 
-        # Start text off-screen to the right
+        self._draw_text_with_outline(draw, text, (0, self.display_height // 2 - 6),
-        x_position = self.display_width
-        
-        # Draw text
-        self._draw_text_with_outline(draw, text, (x_position, self.display_height // 2 - 6), 
                                    font, fill=text_color)
 
         return img
@@ -214,15 +212,9 @@ class DisplayHelper:
         Returns:
             PIL Image with error message
         """
-        img = self.create_base_image((50, 0, 0))  # Dark red background
+        # Dark red background, white text
-
-        # Use default font
         font = ImageFont.load_default()
-        
+        return self.draw_centered_text(message, font, (50, 0, 0), (255, 255, 255))
-        # Draw centered error message
-        self._draw_centered_text(message, font, (50, 0, 0), (255, 255, 255))
-        
-        return img
     
     def draw_no_data_message(self, message: str = "No Data") -> Image.Image:
         """
@@ -234,11 +226,8 @@ class DisplayHelper:
         Returns:
             PIL Image with no data message
         """
-        img = self.create_base_image((0, 0, 0))
         font = ImageFont.load_default()
-        self._draw_centered_text(message, font, (0, 0, 0), (150, 150, 150))
+        return self.draw_centered_text(message, font, (0, 0, 0), (150, 150, 150))
-        
-        return img
     
     def get_display_dimensions(self) -> Tuple[int, int]:
         """

src/common/logo_helper.py

@@ -6,6 +6,8 @@ Extracted from LEDMatrix core to provide reusable functionality for plugins.
 """
 
 import logging
+import os
+import tempfile
 from pathlib import Path
 from typing import Dict, List, Optional, Union
 
@@ -19,6 +21,10 @@ from src.common.permission_utils import (
 )
 
 
+# Well above any real team logo; bounds what a remote URL can write to disk.
+MAX_LOGO_BYTES = 10 * 1024 * 1024
+
+
 class LogoHelper:
     """
     Helper class for logo loading, caching, and resizing.
@@ -137,8 +143,10 @@ class LogoHelper:
         """
         logo_path = Path(logo_path)
         
-        # Try to load existing logo first
+        # Try to load existing logo first. A placeholder written by a previous
-        if logo_path.exists():
+        # failed download does not count: it wears the real logo's filename, so
+        # trusting the file's existence is what left teams as grey boxes.
+        if logo_path.exists() and not self._is_stale_placeholder(logo_path):
             return self.load_logo(team_abbr, logo_path, max_width, max_height)
         
         # Download if URL provided and file doesn't exist
@@ -146,13 +154,61 @@ class LogoHelper:
             try:
                 self.logger.info(f"Downloading logo for {team_abbr} from {logo_url}")
                 self._download_logo(logo_url, logo_path)
+                # The file on disk just changed. Any cached image for it is the
+                # placeholder we came here to replace, and load_logo() answers
+                # from the cache before touching the disk -- so without this the
+                # real logo would not appear until the process restarted.
+                self._invalidate_cached_logo(team_abbr, logo_path)
                 return self.load_logo(team_abbr, logo_path, max_width, max_height)
             except Exception as e:
                 self.logger.error(f"Failed to download logo for {team_abbr}: {e}")
+                # The retry failed, so restart the back-off. The stale
+                # placeholder is still on disk with its old timestamp, and
+                # leaving it there means the next call retries immediately --
+                # a download attempt per call, which is what the back-off
+                # exists to prevent.
+                self._refresh_stale_placeholder(logo_path)
         
         # Create placeholder if all else fails
         return self._create_placeholder_logo(team_abbr, max_width, max_height)
     
+    def _invalidate_cached_logo(self, team_abbr: str, logo_path: Path) -> None:
+        """Drop every cached size of one logo after its file changed on disk."""
+        prefix = f"{team_abbr}_{logo_path}_"
+        for key in [k for k in self._logo_cache if k.startswith(prefix)]:
+            self._logo_cache.pop(key, None)
+            if key in self._cache_order:
+                self._cache_order.remove(key)
+
+    @staticmethod
+    def _refresh_stale_placeholder(logo_path: Path) -> None:
+        """Restart the retry back-off after a failed download attempt."""
+        try:
+            from src.logo_downloader import refresh_placeholder_timestamp
+        except ImportError:
+            return
+        refresh_placeholder_timestamp(logo_path)
+
+    @staticmethod
+    def _is_stale_placeholder(logo_path: Path) -> bool:
+        """True if the file is a placeholder old enough to be worth retrying.
+
+        Imported lazily so this module keeps working against a core build whose
+        logo_downloader predates placeholder marking.
+        """
+        try:
+            from src.logo_downloader import (
+                PLACEHOLDER_RETRY_SECONDS,
+                is_placeholder_logo,
+                placeholder_age_seconds,
+            )
+        except ImportError:
+            return False
+        if not is_placeholder_logo(logo_path):
+            return False
+        age = placeholder_age_seconds(logo_path)
+        return age is None or age >= PLACEHOLDER_RETRY_SECONDS
+
     def get_logo_variations(self, team_abbr: str) -> List[str]:
         """
         Get possible filename variations for a team abbreviation.
@@ -226,7 +282,10 @@ class LogoHelper:
         return {
             'cached_logos': len(self._logo_cache),
             'cache_size_limit': self.cache_size,
-            'cache_usage_percent': (len(self._logo_cache) / self.cache_size) * 100
+            'cache_usage_percent': (
+                (len(self._logo_cache) / self.cache_size) * 100
+                if self.cache_size else 0
+            ),
         }
     
     def _resize_logo(self, logo: Image.Image, max_width: Optional[int] = None, 
@@ -258,17 +317,60 @@ class LogoHelper:
         self._cache_order.append(cache_key)
     
     def _download_logo(self, url: str, file_path: Path) -> None:
-        """Download logo from URL."""
+        """Download logo from URL.
+
+        The response size is capped and the saved file is verified as a
+        decodable image before it is left on disk: a logo URL is remote
+        input, and without this an oversized or malformed response would
+        be cached for every later load_logo() call to trip over.
+
+        The body is streamed and counted as it arrives rather than read
+        through response.content, which buffers the whole thing first —
+        a server that omits Content-Length and never stops sending would
+        exhaust memory before any size check could run. Nothing lands at
+        file_path until the download completes and decodes, so a failed
+        download cannot leave a truncated logo behind either.
+        """
         # Ensure directory exists with proper permissions
         ensure_directory_permissions(file_path.parent, get_assets_dir_mode())
 
-        # Download with timeout
+        # A unique temp name, not a fixed "<name>.part": two plugins can
-        response = self.session.get(url, timeout=30)
+        # ask for the same logo at once, and a shared name would let them
+        # interleave writes into one file, publish the mixture, or delete
+        # each other's partial. Same directory, so os.replace stays atomic.
+        fd, tmp_name = tempfile.mkstemp(
+            dir=str(file_path.parent), prefix=file_path.name + '.', suffix='.part')
+        tmp_path = Path(tmp_name)
+        try:
+            # fdopen outermost so the descriptor mkstemp handed back is
+            # always adopted and closed, including when the request itself
+            # raises — load_logo_with_download swallows that, so a leak
+            # here would accumulate quietly on a URL that keeps failing.
+            with os.fdopen(fd, 'wb') as f:
+                with self.session.get(url, timeout=30, stream=True) as response:
                     response.raise_for_status()
+                    downloaded = 0
+                    for chunk in response.iter_content(chunk_size=64 * 1024):
+                        if not chunk:
+                            continue
+                        downloaded += len(chunk)
+                        if downloaded > MAX_LOGO_BYTES:
+                            raise ValueError(
+                                f"Logo at {url} exceeds the "
+                                f"{MAX_LOGO_BYTES}-byte limit; not saved")
+                        f.write(chunk)
 
-        # Save to file
+            # Verify it decodes before it becomes the cached logo. PIL
-        with open(file_path, 'wb') as f:
+            # raises DecompressionBombError past its own pixel limit; a
-            f.write(response.content)
+            # partial or non-image response raises UnidentifiedImageError
+            # (an OSError subclass).
+            with Image.open(tmp_path) as probe:
+                probe.load()
+
+            os.replace(tmp_path, file_path)
+        except BaseException:
+            tmp_path.unlink(missing_ok=True)
+            raise
 
         # Set proper file permissions after saving
         ensure_file_permissions(file_path, get_assets_file_mode())

src/common/scroll_helper.py

@@ -328,7 +328,11 @@ class ScrollHelper:
             elapsed_time = current_time - (self.scroll_start_time or current_time)
             # The image already includes display_width padding, so we only need total_scroll_width
             required_total_distance = self.total_scroll_width
-            self.logger.info(
+            # Progress telemetry, emitted every few seconds for the whole of
+            # every scroll. It says how far along a marquee is, which is what
+            # you turn debug on to watch and not something an operator needs
+            # in the journal on a device that scrolls all day.
+            self.logger.debug(
                 "Scroll progress: elapsed=%.2fs, target=%.2fs, total_scrolled=%.0f/%d px (%.1f%%)",
                 elapsed_time,
                 self.calculated_duration,

src/common/sports_card.py

@@ -0,0 +1,470 @@
+"""Card-drawing helpers shared by every sports scoreboard plugin.
+
+The eight scoreboards each carried byte-identical copies of the functions
+below: the colour pickers, the settings lookup, the date and time formatting,
+the favourite-team rules and the font-size grid snapping. One fix had to be
+made eight times, and a new scoreboard started by copying them a ninth.
+
+Everything here is a **free function taking explicit arguments**, not a base
+class. Adoption is therefore per-function and reversible: a plugin keeps its
+method and delegates the body, so the call sites and the override points are
+untouched. That is also why `config`, `logger` and `fonts` are parameters
+rather than attributes -- the helper never reaches back into the caller.
+
+The bodies are the plugins' own code, moved rather than rewritten. The one
+deliberate difference is `crisp_size`, which takes the seven-plugin guard
+(`not desired`) instead of football's: they agree on every real input, and
+the extra guard only stops a None size raising TypeError.
+"""
+
+import logging
+from datetime import datetime, timezone
+from typing import Any, Dict, Optional, Tuple
+from zoneinfo import ZoneInfo
+
+logger = logging.getLogger(__name__)
+
+__all__ = [
+    "ELEMENT_FOR_FONT", "FAVORITE_RESULT_COLOR_DEFAULTS", "FONT_NAME_ALIASES",
+    "FONT_PIXEL_GRID", "MONTH_ABBR", "WEEKDAY_ABBR",
+    "scroll_card_option", "element_color", "font_color", "coerce_rgb",
+    "score_color_for", "recent_score_color", "favorite_teams_for",
+    "side_is_favorite", "side_score", "favorite_result",
+    "card_tzinfo", "weekday_for", "format_game_date", "format_game_time",
+    "vs_text", "upcoming_center_mode", "crisp_size", "schema_font_size",
+    "resolve_font_size", "unshare_element_fonts",
+]
+
+#: Which customization element owns each font key, for colour resolution.
+ELEMENT_FOR_FONT: Dict[str, str] = {
+    "score": "score_text",
+    "time": "period_text",
+    "team": "team_name",
+    "status": "status_text",
+    "detail": "detail_text",
+    "rank": "rank_text",
+}
+
+#: Fallback colours when favourite_result_colors is on but a slot is unset.
+FAVORITE_RESULT_COLOR_DEFAULTS: Dict[str, Tuple[int, int, int]] = {
+    "win": (0, 255, 0),
+    "loss": (255, 0, 0),
+    "tie": (255, 200, 0),
+}
+
+#: Family aliases the web UI may write, mapped to the shipped filename.
+FONT_NAME_ALIASES: Dict[str, str] = {
+    "press_start": "PressStart2P-Regular.ttf",
+    "four_by_six": "4x6-font.ttf",
+}
+
+#: Pixel grid each face renders crisply on. Off-grid sizes anti-alias, which
+#: on an LED matrix is a dim lamp rather than a soft edge.
+FONT_PIXEL_GRID: Dict[str, int] = {
+    "PressStart2P-Regular.ttf": 8,
+    "4x6-font.ttf": 7,
+}
+
+MONTH_ABBR = ("Jan", "Feb", "Mar", "Apr", "May", "Jun",
+              "Jul", "Aug", "Sep", "Oct", "Nov", "Dec")
+WEEKDAY_ABBR = ("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun")
+
+
+# ---------------------------------------------------------------------------
+# Settings lookup
+# ---------------------------------------------------------------------------
+
+def scroll_card_option(config: Optional[Dict[str, Any]], key: str,
+                       default: Any = None) -> Any:
+    """Read one key from the scroll_card config block."""
+    block = (config or {}).get("scroll_card")
+    if isinstance(block, dict) and block.get(key) is not None:
+        return block.get(key)
+    return default
+
+
+def vs_text(config: Optional[Dict[str, Any]]) -> str:
+    """Separator drawn between the teams -- "VS", "@", "at", anything."""
+    return str(scroll_card_option(config, "vs_text", "VS"))
+
+
+def upcoming_center_mode(config: Optional[Dict[str, Any]]) -> str:
+    """Middle of an upcoming card: 'vs', 'date_time' or 'none'."""
+    mode = str(scroll_card_option(config, "upcoming_center", "vs") or "vs").lower()
+    return mode if mode in ("vs", "date_time", "none") else "vs"
+
+
+# ---------------------------------------------------------------------------
+# Colour
+# ---------------------------------------------------------------------------
+
+def element_color(config: Optional[Dict[str, Any]], element: str,
+                  default: Tuple[int, int, int] = (255, 255, 255)):
+    """Per-element text colour from customization.<element>.text_color."""
+    try:
+        cfg = (config or {}).get("customization", {}).get(element, {})
+        value = cfg.get("text_color")
+        if isinstance(value, (list, tuple)) and len(value) == 3:
+            return tuple(max(0, min(255, int(c))) for c in value)
+        if isinstance(value, str) and value.startswith("#") and len(value) == 7:
+            return tuple(int(value[i:i + 2], 16) for i in (1, 3, 5))
+    except (TypeError, ValueError):
+        pass
+    return default
+
+
+def font_color(config: Optional[Dict[str, Any]], fonts: Optional[Dict[str, Any]],
+               font, default: Tuple[int, int, int] = (255, 255, 255)):
+    """Colour for whichever element owns this face.
+
+    Matched on identity, and deliberately gives up when one object is
+    shared: the last-resort font path can hand the same face to several
+    keys, and there is no right answer for which element's colour that is.
+    White is what those draws used before, so ambiguity costs nothing.
+    """
+    try:
+        fonts = fonts or {}
+        matches = [element for key, element in ELEMENT_FOR_FONT.items()
+                   if fonts.get(key) is font]
+        if len(matches) == 1:
+            return element_color(config, matches[0], default)
+    except (AttributeError, TypeError):
+        pass
+    return default
+
+
+def coerce_rgb(value, fallback):
+    """Turn a configured [R, G, B] list into a clamped (r, g, b) tuple."""
+    # Checked before unpacking: a 3-character string ("123") would otherwise
+    # iterate into three digits and yield a colour rather than the fallback.
+    if not isinstance(value, (list, tuple)) or len(value) != 3:
+        return fallback
+    try:
+        r, g, b = (max(0, min(255, int(channel))) for channel in value)
+    except (TypeError, ValueError):
+        return fallback
+    return (r, g, b)
+
+
+# ---------------------------------------------------------------------------
+# Favourite teams
+# ---------------------------------------------------------------------------
+
+def favorite_teams_for(config: Dict[str, Any], game: Dict[str, Any]) -> list:
+    """Favorite teams that apply to this game.
+
+    Both sources are used. Games carry the league manager's *resolved*
+    favorites, which is the only place dynamic groups such as AP_TOP_25
+    appear expanded; the config is read as well so an edit takes effect on
+    already-fetched games, and so hand-built game dicts (tests, other
+    callers) still work.
+    """
+    favorites = list(game.get("favorite_teams") or [])
+    league_config = config.get(str(game.get("league", "") or ""))
+    if isinstance(league_config, dict):
+        favorites += list(league_config.get("favorite_teams") or [])
+    else:
+        favorites += list(config.get("favorite_teams") or [])
+    return favorites
+
+
+def side_is_favorite(game: Dict[str, Any], side: str, favorites: set) -> bool:
+    """Is the home/away side of this game a favorite team?
+
+    Reads both the flat (``home_abbr``) and nested (``home_team.abbrev``)
+    payload shapes, and matches on the ESPN id too, because a couple of
+    leagues (NRL) key favorites by id where abbreviations collide.
+    """
+    candidates = [game.get(f"{side}_abbr"), game.get(f"{side}_id")]
+    team = game.get(f"{side}_team")
+    if isinstance(team, dict):
+        candidates += [team.get("abbrev"), team.get("abbreviation"), team.get("id")]
+    for value in candidates:
+        if value is not None and str(value).strip().upper() in favorites:
+            return True
+    return False
+
+
+def side_score(game: Dict[str, Any], side: str) -> Optional[int]:
+    """Numeric score for one side, from either payload shape."""
+    raw = None
+    team = game.get(f"{side}_team")
+    if isinstance(team, dict) and team.get("score") is not None:
+        raw = team.get("score")
+    if raw is None:
+        raw = game.get(f"{side}_score")
+    try:
+        return int(float(str(raw).strip()))
+    except (TypeError, ValueError):
+        return None
+
+
+def favorite_result(config: Dict[str, Any], game: Dict[str, Any]) -> Optional[str]:
+    """Say how the favorite team did in a finished game.
+
+    Returns 'win', 'loss' or 'tie', or None when there is no single team
+    to root for: no favorites configured, neither side is a favorite, or
+    *both* are -- a favorite-vs-favorite game has no losing side worth
+    flagging in red. Also None when the scores are not usable numbers.
+    """
+    favorites = {
+        str(team).strip().upper()
+        for team in favorite_teams_for(config, game)
+        if str(team).strip()
+    }
+    if not favorites:
+        return None
+
+    home_fav = side_is_favorite(game, "home", favorites)
+    away_fav = side_is_favorite(game, "away", favorites)
+    if home_fav == away_fav:
+        return None
+
+    home_score = side_score(game, "home")
+    away_score = side_score(game, "away")
+    if home_score is None or away_score is None:
+        return None
+
+    if home_score == away_score:
+        return "tie"
+    favorite_score, other_score = (
+        (home_score, away_score) if home_fav else (away_score, home_score)
+    )
+    return "win" if favorite_score > other_score else "loss"
+
+
+def recent_score_color(config: Dict[str, Any], logger, game: Dict[str, Any], default):
+    """Fill color for a finished game's score, per favorite_result_colors."""
+    try:
+        settings = (config.get("customization") or {}).get(
+            "favorite_result_colors"
+        ) or {}
+        if not settings.get("enabled", False):
+            return default
+        result = favorite_result(config, game)
+        if result is None:
+            return default
+        return coerce_rgb(
+            settings.get(f"{result}_color"),
+            FAVORITE_RESULT_COLOR_DEFAULTS[result],
+        )
+    except Exception:
+        logger.debug("Could not resolve favorite result color", exc_info=True)
+        return default
+
+
+def score_color_for(config: Dict[str, Any], logger, game: Dict[str, Any],
+                    game_type: str, default=None):
+    """Fill color for a game card's score. Only finished games are tinted.
+
+    The default is the configured score colour rather than a flat white,
+    so customization.score_text.text_color shows on games the favourite
+    tint does not apply to. The tint still wins where it applies.
+    """
+    if default is None:
+        default = element_color(config, 'score_text')
+    if game_type != "recent":
+        return default
+    return recent_score_color(config, logger, game, default)
+
+
+# ---------------------------------------------------------------------------
+# Date and time
+# ---------------------------------------------------------------------------
+
+def card_tzinfo(config: Optional[Dict[str, Any]], logger):
+    """Timezone for weekday/24h conversions; falls back to UTC."""
+    configured = (config or {}).get("timezone")
+    if configured:
+        try:
+            return ZoneInfo(configured)
+        except (KeyError, ValueError, TypeError, OSError) as exc:
+            # KeyError covers ZoneInfoNotFoundError. A bad zone name in
+            # config should fall back to UTC, not blank the card.
+            logger.debug("Unusable timezone %r: %s", configured, exc)
+    return timezone.utc
+
+
+def weekday_for(config: Optional[Dict[str, Any]], logger,
+                game: Optional[Dict]) -> str:
+    """Weekday abbreviation from the game's start time, or ''."""
+    if not game:
+        return ""
+    raw = game.get("start_time_utc") or game.get("start_time")
+    if not raw:
+        return ""
+    try:
+        start = raw if isinstance(raw, datetime) else datetime.fromisoformat(
+            str(raw).replace("Z", "+00:00"))
+        return WEEKDAY_ABBR[start.astimezone(card_tzinfo(config, logger)).weekday()]
+    except (ValueError, TypeError):
+        return ""
+
+
+def format_game_date(config: Optional[Dict[str, Any]], logger, date_text: str,
+                     game: Optional[Dict] = None) -> str:
+    """Format an upcoming card's date per scroll_card.date_format."""
+    raw = str(date_text or "").strip()
+    if not raw:
+        return ""
+    fmt = str(scroll_card_option(config, "date_format", "abbrev") or "abbrev")
+    if fmt == "numeric":
+        return raw
+    parts = raw.replace("-", "/").split("/")
+    if not (len(parts) >= 2 and parts[0].strip().isdigit() and parts[1].strip().isdigit()):
+        return raw
+    month, day = int(parts[0]), int(parts[1])
+    if not 1 <= month <= 12:
+        return raw
+    name = MONTH_ABBR[month - 1]
+    if fmt == "numeric_day_first":
+        return f"{day}/{month}"
+    if fmt == "day_first":
+        return f"{day} {name}"
+    if fmt == "weekday":
+        weekday = weekday_for(config, logger, game)
+        return f"{weekday} {name} {day}" if weekday else f"{name} {day}"
+    return f"{name} {day}"
+
+
+def format_game_time(config: Optional[Dict[str, Any]], time_text: str) -> str:
+    """Return the time as-is (12h) or converted to 24h."""
+    raw = str(time_text or "").strip()
+    if not raw or str(scroll_card_option(config, "time_format", "12h")) != "24h":
+        return raw
+    cleaned = raw.upper().replace(" ", "")
+    meridiem = "AM" if cleaned.endswith("AM") else "PM" if cleaned.endswith("PM") else ""
+    if not meridiem:
+        return raw
+    try:
+        hh, _, mm = cleaned[:-2].partition(":")
+        hour, minute = int(hh), int(mm or 0)
+    except ValueError:
+        return raw
+    if not (0 <= hour <= 12 and 0 <= minute <= 59):
+        return raw
+    hour = hour % 12 + (12 if meridiem == "PM" else 0)
+    return f"{hour:02d}:{minute:02d}"
+
+
+# ---------------------------------------------------------------------------
+# Font sizing
+# ---------------------------------------------------------------------------
+
+#: Per-schema caches, keyed by the schema's absolute path. Keyed rather than
+#: global because each plugin declares its own defaults; keyed rather than
+#: per-class because the helper has no class to hang it on.
+_SCHEMA_FONT_SIZE_CACHE: Dict[str, Dict[str, int]] = {}
+
+
+def crisp_size(font_file, desired, aliases=None, grid_table=None):
+    """Snap *desired* to the nearest size *font_file* renders crisply at.
+
+    A face with no known grid is returned unchanged, so a user-supplied
+    font is never second-guessed.
+
+    ``aliases`` and ``grid_table`` default to the shared tables; a plugin
+    that ships an extra face can pass its own without forking this.
+    """
+    aliases = FONT_NAME_ALIASES if aliases is None else aliases
+    grid_table = FONT_PIXEL_GRID if grid_table is None else grid_table
+    font_file = aliases.get(font_file, font_file)
+    grid = grid_table.get(font_file)
+    if not grid or not desired or desired <= 0:
+        return desired
+    return max(grid, int(round(float(desired) / grid)) * grid)
+
+
+def schema_font_size(schema_path: str, element_key) -> Optional[int]:
+    """The font_size this plugin's config_schema.json declares, or None.
+
+    Cached per schema path. The plugins cached this on their own class; the
+    path is the same distinction expressed without one, so two plugins never
+    share an entry.
+    """
+    if not element_key:
+        return None
+    cache = _SCHEMA_FONT_SIZE_CACHE.get(schema_path)
+    if cache is None:
+        cache = {}
+        try:
+            import json
+            with open(schema_path) as fh:
+                schema = json.load(fh)
+            props = (schema.get('properties', {})
+                           .get('customization', {})
+                           .get('properties', {}))
+            for key, spec in props.items():
+                size = spec.get('properties', {}).get('font_size', {}).get('default')
+                if size is not None:
+                    cache[key] = int(size)
+        except Exception as exc:
+            # See sports_shared._schema_font_size: an unreadable schema
+            # silently disables the pixel-grid snap for every element.
+            # Built once per schema path, so this cannot repeat per frame.
+            logger.warning(
+                "could not read %s (%s: %s); font sizes will skip their "
+                "pixel grid snap and may render a pixel narrow",
+                schema_path, type(exc).__name__, exc)
+            cache = {}
+        _SCHEMA_FONT_SIZE_CACHE[schema_path] = cache
+    return cache.get(element_key)
+
+
+def resolve_font_size(schema_path: str, element_config, element_key,
+                      default_size, font_name, aliases=None, grid_table=None):
+    """Size to render at: the user's choice, or a grid-snapped default.
+
+    A configured size counts as a real choice only when it differs from
+    the schema default. The web UI writes the whole schema default block
+    on every save, so "font_size == schema default" carries no intent and
+    would otherwise pin every install to an anti-aliased size forever.
+    """
+    configured = (element_config or {}).get('font_size')
+    if configured is not None:
+        try:
+            configured = int(configured)
+            if configured != schema_font_size(schema_path, element_key):
+                return configured
+        except (TypeError, ValueError):
+            pass
+    return crisp_size(font_name, default_size, aliases, grid_table)
+
+
+def unshare_element_fonts(logger, fonts):
+    """Give each colourable element its own face object.
+
+    The colour a draw gets is resolved from the face it was handed, and
+    several of these loaders legitimately hand one object to more than one
+    element -- a size resolver that lands two elements on the same face, a
+    fallback that fills every key from one default, football's narrowing
+    step that deliberately shrinks the clock along with the score. Sharing
+    the object makes the element ambiguous and the colour unresolvable.
+
+    Re-instantiating from the same path and size gives a distinct object
+    with identical metrics, so nothing about the rendering changes; only
+    the ability to tell two elements apart does. Faces that cannot be
+    rebuilt (a BDF loaded through freetype.Face, anything without a usable
+    path) are left shared, and their draws stay white as before.
+    """
+    try:
+        from PIL import ImageFont as _IF
+    except ImportError:  # pragma: no cover
+        return fonts
+    seen = {}
+    for key in ELEMENT_FOR_FONT:
+        font = fonts.get(key)
+        if font is None:
+            continue
+        if id(font) not in seen:
+            seen[id(font)] = key
+            continue
+        path, size = getattr(font, "path", None), getattr(font, "size", None)
+        if not path or not size:
+            continue
+        try:
+            fonts[key] = _IF.truetype(path, size)
+        except (OSError, ValueError, TypeError):
+            logger.debug(
+                "Could not un-share the %s face; it keeps the default colour", key)
+    return fonts

src/common/sports_game_renderer.py

@@ -0,0 +1,261 @@
+"""The scroll/Vegas card geometry the sports scoreboards all share.
+
+Eight scoreboards -- afl, baseball, basketball, football, hockey, lacrosse,
+nrl and soccer -- each carried their own ``game_renderer.py``. After the card
+helpers moved to ``src/common/sports_card.py`` the settings lookups were
+shared, but the *geometry* was not: nine methods that decide how wide the
+centre gap is, how much room a logo gets, and where an upcoming card's date
+and time land were still eight separate copies. Five were byte-identical
+across all eight plugins; the other four were identical in seven, with a
+different single outlier each time.
+
+That last detail is why this is a mixin rather than free functions. There is
+no per-sport branching to write -- baseball needs its own
+``_draw_upcoming_game_status`` and ``_logo_slot_width``, hockey its own
+``_upcoming_date_and_time``, football its own ``_score_reserve_width``, and
+every one of those is an ordinary override. The seven that agree inherit and
+say nothing.
+
+It is deliberately *only* a mixin: no ``__init__``, no state of its own. The
+plugins' constructors differ in six ways and none of that difference is worth
+unifying, so adoption is one line on the class statement plus deleting the
+methods that now come from here.
+
+What a host class must provide
+------------------------------
+Attributes: ``display_width``, ``display_height``, ``config``, ``fonts``,
+``logger``, ``_team_rankings_cache``.
+
+Methods: ``_draw_text_with_outline(draw, text, position, font, fill=None,
+outline_color=(0, 0, 0))`` -- the one hook whose body genuinely varies -- plus
+the ``sports_card`` delegations ``_scroll_card_option``,
+``_upcoming_center_mode``, ``_vs_text``, ``_element_color``,
+``_format_game_date`` and ``_format_game_time``.
+"""
+
+import math
+from typing import ClassVar, Dict, Tuple
+
+from PIL import Image, ImageDraw
+
+
+class SportsGameRendererMixin:
+    """Shared card geometry for the sports scoreboards. See module docstring."""
+
+    #: Centre strip as a fraction of card width, before clamping.
+    CENTER_GAP_RATIO: ClassVar[float] = 0.28
+    #: Clamp floor for the derived centre gap.
+    CENTER_GAP_MIN_PX: ClassVar[int] = 22
+    #: Clamp ceiling for the derived centre gap.
+    CENTER_GAP_MAX_PX: ClassVar[int] = 40
+    #: Breathing room kept between the score and each logo.
+    _SCORE_LOGO_GUTTER_PX: ClassVar[int] = 4
+    #: Widest score the centre strip must fit. Leagues that can reach three
+    #: digits a side override this with "000-000".
+    _SCORE_PROBE: ClassVar[str] = "00-00"
+
+    # ---- geometry ------------------------------------------------------
+
+    # Non-finite settings are rejected before any int()/round(): "inf" reaches
+    # these from config as a float or a string, passes an `isinstance` plus
+    # `>= 0` check unharmed, and then raises OverflowError out of int() --
+    # which the old `except (TypeError, ValueError)` did not catch, so it
+    # aborted the whole card render. Present in all eight plugins before this
+    # moved to the core; fixing it here fixes it in all eight.
+
+    def _score_reserve_width(self) -> int:
+        """Centre strip the score actually needs, measured rather than assumed.
+
+        The gap was derived from the card width alone (width x
+        CENTER_GAP_RATIO, clamped to CENTER_GAP_MAX_PX) while the score's size
+        comes from config and the element-style resolver. Nothing compared the
+        two, so any score wider than the clamp was drawn over the logos.
+        Measuring it keeps the strip wide enough for whatever font is in play.
+        """
+        try:
+            probe = ImageDraw.Draw(Image.new("RGB", (4, 4)))
+            width = probe.textlength(self._SCORE_PROBE, font=self.fonts['score'])
+            return int(width) + 2 * self._SCORE_LOGO_GUTTER_PX
+        except Exception:
+            self.logger.debug("Score reserve measurement failed", exc_info=True)
+            return 0
+
+    def _center_gap_width(self) -> int:
+        """Width of the middle strip kept clear of logos.
+
+        ``scroll_card.center_gap`` pins it outright; otherwise it scales with
+        the card width between the configurable min and max. 0 restores
+        edge-to-edge logos.
+        """
+        configured = self._scroll_card_option("center_gap")
+        if (isinstance(configured, (int, float))
+                and math.isfinite(configured) and configured >= 0):
+            return int(configured)
+        ratio = self._scroll_card_option("center_gap_ratio", self.CENTER_GAP_RATIO)
+        low = self._scroll_card_option("center_gap_min", self.CENTER_GAP_MIN_PX)
+        high = self._scroll_card_option("center_gap_max", self.CENTER_GAP_MAX_PX)
+        try:
+            ratio, low, high = float(ratio), float(low), float(high)
+            if not (math.isfinite(ratio) and math.isfinite(low)
+                    and math.isfinite(high)):
+                return self.CENTER_GAP_MIN_PX
+            scaled = round(self.display_width * ratio)
+            derived = int(max(int(low), min(int(high), scaled)))
+            # A strip narrower than the score is the bug, not a style choice.
+            # An explicit ``center_gap`` is still honoured above, including 0.
+            return max(derived, self._score_reserve_width())
+        except (TypeError, ValueError, OverflowError):
+            return self.CENTER_GAP_MIN_PX
+
+    def _logo_slot_width(self) -> int:
+        """Per-side logo slot, leaving the center gap clear.
+
+        No longer capped at display_height: the card is sized as two
+        full-height logos plus the measured gap, so what is left after the gap
+        is exactly the logo's share. The cap was what froze the logos at 46px
+        on the old flat 128px card.
+        """
+        available = (self.display_width - self._center_gap_width()) // 2
+        return max(8, available)
+
+    def _logo_cache_key(self, name: str) -> str:
+        """Cache key scoped to the logo slot.
+
+        One cache dict is shared by renderers built for different card widths,
+        so a logo sized for a wide slot must not be handed to a narrow one.
+        """
+        return f"{name}@{self._logo_slot_width()}x{self.display_height}"
+
+    def _layout_offset(self, element: str, axis: str, default: int = 0) -> int:
+        """X/Y nudge for one element, from customization.layout.
+
+        Same block the full-screen scorebug reads (sports.py
+        _get_layout_offset), so a nudge configured in the web UI now moves
+        the element on the scroll/Vegas card too -- previously the schema
+        advertised these offsets but this renderer ignored them.
+        """
+        try:
+            layout = (self.config or {}).get("customization", {}).get("layout", {})
+            value = (layout.get(element) or {}).get(axis, default)
+            if isinstance(value, bool):
+                return default
+            if isinstance(value, (int, float)):
+                return int(value) if math.isfinite(value) else default
+            if isinstance(value, str):
+                parsed = float(value)
+                return int(parsed) if math.isfinite(parsed) else default
+        except (TypeError, ValueError, OverflowError):
+            pass
+        return default
+
+    # ---- upcoming cards ------------------------------------------------
+
+    def _upcoming_date_and_time(self, game: Dict) -> Tuple[str, str]:
+        """(date, time) for an upcoming card, from the extractor's flat keys."""
+        return (
+            str(game.get("game_date", "") or ""),
+            str(game.get("game_time", "") or ""),
+        )
+
+    def _draw_upcoming_center(self, draw: "ImageDraw.ImageDraw", game: Dict) -> None:
+        """Draw the middle of an upcoming card.
+
+        Never a score: an upcoming game has not started, so the extractor's
+        0-0 is noise. Either the VS text (default), the date and time stacked,
+        or nothing at all.
+        """
+        mode = self._upcoming_center_mode()
+        if mode == "none":
+            return
+
+        if mode == "vs":
+            vs_text = self._vs_text()
+            if not vs_text:
+                return
+            vs_width = draw.textlength(vs_text, font=self.fonts['score'])
+            vs_x = (self.display_width - vs_width) // 2 + self._layout_offset('score', 'x_offset')
+            vs_y = (self.display_height // 2) - 3 + self._layout_offset('score', 'y_offset')
+            self._draw_text_with_outline(
+                draw, vs_text, (vs_x, vs_y), self.fonts['score'],
+                fill=self._element_color('score_text')
+            )
+            return
+
+        date_text, time_text = self._upcoming_date_and_time(game)
+        lines = []
+        if self._scroll_card_option("show_date", True):
+            lines.append(self._format_game_date(date_text, game))
+        if self._scroll_card_option("show_time", True):
+            lines.append(self._format_game_time(time_text))
+        lines = [t for t in lines if t]
+        if not lines:
+            return
+        font = self.fonts.get('detail') or self.fonts['time']
+        line_h = 7
+        top = (self.display_height // 2) - (len(lines) * line_h) // 2
+        top += self._layout_offset('score', 'y_offset')
+        for i, line in enumerate(lines):
+            width = draw.textlength(line, font=font)
+            x = (self.display_width - width) // 2 + self._layout_offset('score', 'x_offset')
+            self._draw_text_with_outline(
+                draw, line, (x, top + i * line_h), font,
+                fill=self._element_color('detail_text')
+            )
+
+    def _draw_upcoming_game_status(self, draw: "ImageDraw.ImageDraw", game: Dict) -> None:
+        """Draw the date and time around an upcoming card.
+
+        Time top and date bottom by default; scroll_card.swap_date_time puts
+        the date on top instead. Skipped when the pair is stacked in the
+        middle, which would otherwise print them twice.
+        """
+        if self._upcoming_center_mode() == "date_time":
+            return
+
+        date_raw, time_raw = self._upcoming_date_and_time(game)
+        date_text = (self._format_game_date(date_raw, game)
+                     if self._scroll_card_option("show_date", True) else "")
+        time_text = (self._format_game_time(time_raw)
+                     if self._scroll_card_option("show_time", True) else "")
+
+        if self._scroll_card_option("swap_date_time", False):
+            top_text, top_el, bottom_text, bottom_el = (
+                date_text, 'date', time_text, 'time')
+            top_font = self.fonts.get('detail') or self.fonts['time']
+            bottom_font = self.fonts['time']
+            top_color, bottom_color = 'detail_text', 'period_text'
+        else:
+            top_text, top_el, bottom_text, bottom_el = (
+                time_text, 'time', date_text, 'date')
+            top_font = self.fonts['time']
+            bottom_font = self.fonts.get('detail') or self.fonts['time']
+            top_color, bottom_color = 'period_text', 'detail_text'
+
+        if top_text:
+            top_width = draw.textlength(top_text, font=top_font)
+            top_x = (self.display_width - top_width) // 2 + self._layout_offset(top_el, 'x_offset')
+            top_y = 1 + self._layout_offset(top_el, 'y_offset')
+            self._draw_text_with_outline(
+                draw, top_text, (top_x, top_y), top_font,
+                fill=self._element_color(top_color)
+            )
+
+        if bottom_text:
+            bottom_width = draw.textlength(bottom_text, font=bottom_font)
+            bottom_x = ((self.display_width - bottom_width) // 2
+                        + self._layout_offset(bottom_el, 'x_offset'))
+            # Measured, not a fixed -7: the detail font is 6px in most plugins
+            # but 10px in soccer and nrl, where "Sep 19" ran past the card.
+            ink_bottom = draw.textbbox((0, 0), bottom_text, font=bottom_font)[3]
+            bottom_y = (max(0, self.display_height - ink_bottom - 1)
+                        + self._layout_offset(bottom_el, 'y_offset'))
+            self._draw_text_with_outline(
+                draw, bottom_text, (bottom_x, bottom_y), bottom_font,
+                fill=self._element_color(bottom_color)
+            )
+
+    # ---- rankings ------------------------------------------------------
+
+    def set_rankings_cache(self, rankings: Dict[str, int]) -> None:
+        """Set the team rankings cache for display."""
+        self._team_rankings_cache = rankings

src/common/sync_manager.py

@@ -19,6 +19,7 @@ Port default: 5765 (UDP). Open this port on both Pis if ufw is active:
 
 import io
 import json
+import math
 import os
 import socket
 import struct
@@ -37,6 +38,13 @@ _RAW_MAGIC = b'SYNC_RAW'
 _RAW_HEADER = struct.Struct('<HH')  # width, height (uint16 LE)
 
 
+# Upper bound on a decoded frame/scroll image. Generous for any real scroll
+# image (a leader's full cycle is long but only panel-height tall), and low
+# enough that a crafted image from any host on the LAN cannot force a large
+# allocation on the render thread. Applied on both receive paths — the TCP
+# image server and the follower's legacy-PNG UDP fallback.
+_MAX_FRAME_W, _MAX_FRAME_H = 100_000, 256
+
 SYNC_PORT = 5765
 HELLO_INTERVAL = 5.0       # follower broadcasts hello every 5 s
 HEARTBEAT_INTERVAL = 2.0   # follower sends heartbeat every 2 s
@@ -101,6 +109,7 @@ class DisplaySyncManager:
         self._peer_chain: int = 0
         self._last_heartbeat_time: float = 0.0
         self._leader_width: int = 0  # set by display_controller after init
+        self._oversized_frame_warned: bool = False
 
         # Follower state
         self._follower_state = FollowerState.STANDALONE
@@ -174,6 +183,10 @@ class DisplaySyncManager:
                 continue
             except Exception as exc:
                 self.logger.debug("Sync leader recv error: %s", exc)
+                # Brief backoff: a socket left in a bad state raises
+                # immediately, which would otherwise spin this thread at
+                # 100% CPU logging the same error.
+                time.sleep(0.1)
 
     def _handle_hello(self, msg: dict, sender_ip: str) -> None:
         hw = self._hw_config
@@ -273,11 +286,10 @@ class DisplaySyncManager:
                             break
                         data.extend(chunk)
                     img = Image.open(io.BytesIO(data))
-                    _MAX_W, _MAX_H = 100_000, 256  # generous for any real scroll image
+                    if img.width > _MAX_FRAME_W or img.height > _MAX_FRAME_H:
-                    if img.width > _MAX_W or img.height > _MAX_H:
                         self.logger.warning(
                             "Sync: rejected oversized scroll image %dx%d (max %dx%d) from %s",
-                            img.width, img.height, _MAX_W, _MAX_H, addr,
+                            img.width, img.height, _MAX_FRAME_W, _MAX_FRAME_H, addr,
                         )
                         continue
                     try:
@@ -396,7 +408,7 @@ class DisplaySyncManager:
             data = header + arr.tobytes()
             if len(data) <= 65000:
                 self._send_sock.sendto(data, (self._peer_ip, self.port))
-            elif not getattr(self, '_oversized_frame_warned', False):
+            elif not self._oversized_frame_warned:
                 self._oversized_frame_warned = True
                 self.logger.warning(
                     "Sync: frame too large for UDP (%d bytes, max 65000) — "
@@ -451,25 +463,8 @@ class DisplaySyncManager:
         )
         self.write_status_file()
 
-    def _follower_recv_loop(self) -> None:
+    def _handle_received_frame(self, img: Image.Image, sender_ip: str) -> None:
-        while self._running:
+        """Record a decoded leader frame and enter follower mode if needed."""
-            try:
-                data, addr = self._recv_sock.recvfrom(65535)
-                sender_ip = addr[0]
-
-                if data[:8] == _RAW_MAGIC or len(data) > 512:
-                    # Frame data: prefer magic-tagged raw RGB; fall back to legacy PNG
-                    try:
-                        if data[:8] == _RAW_MAGIC:
-                            w, h = _RAW_HEADER.unpack(data[8:12])
-                            raw = data[12:]
-                            img = Image.frombuffer(
-                                "RGB", (w, h), raw, "raw", "RGB", 0, 1
-                            )
-                        else:
-                            # Fallback: try legacy PNG
-                            img = Image.open(io.BytesIO(data))
-                            img.load()
         with self._frame_lock:
             self._latest_frame = img
         self._last_leader_frame_time = time.time()
@@ -482,12 +477,62 @@ class DisplaySyncManager:
                 sender_ip,
             )
             self.write_status_file()
+
+    def _follower_recv_loop(self) -> None:
+        while self._running:
+            try:
+                data, addr = self._recv_sock.recvfrom(65535)
+                sender_ip = addr[0]
+
+                if data[:8] == _RAW_MAGIC:
+                    # Magic-tagged raw RGB frame — self-describing, no guessing.
+                    try:
+                        w, h = _RAW_HEADER.unpack(data[8:12])
+                        raw = data[12:]
+                        img = Image.frombuffer(
+                            "RGB", (w, h), raw, "raw", "RGB", 0, 1
+                        )
+                        self._handle_received_frame(img, sender_ip)
                     except Exception as exc:
                         self.logger.debug("Sync: frame decode error: %s", exc)
                 else:
-                    # Control message
+                    # No magic prefix. Whether the payload parses as JSON
+                    # decides between a control message and a legacy
+                    # (pre-magic) PNG frame — both wire formats are
+                    # self-describing, so no size heuristic is needed. A
+                    # >512-byte control message used to be misrouted into
+                    # image decode and silently dropped.
                     try:
                         msg = json.loads(data.decode("utf-8"))
+                    except (json.JSONDecodeError, UnicodeDecodeError):
+                        # Not JSON — try a legacy PNG frame.
+                        try:
+                            img = Image.open(io.BytesIO(data))
+                            if img.width > _MAX_FRAME_W or img.height > _MAX_FRAME_H:
+                                # Same cap the TCP image path applies: decode
+                                # is deferred until load(), so check first.
+                                self.logger.debug(
+                                    "Sync: rejected oversized legacy frame %dx%d from %s",
+                                    img.width, img.height, sender_ip,
+                                )
+                                continue
+                            img.load()
+                            self._handle_received_frame(img, sender_ip)
+                        except Exception as exc:
+                            self.logger.debug("Sync: frame decode error: %s", exc)
+                        continue
+
+                    # It parsed, so it is a control message and never a
+                    # frame. Read and validate its fields under a guard —
+                    # a UDP payload is attacker-shaped, so a non-object
+                    # body makes .get() raise AttributeError and an "sx"
+                    # carrying a non-numeric x raises ValueError/TypeError
+                    # — but dispatch the callback *outside* it. Running
+                    # the callback in here would let a fault in someone
+                    # else's code read as a malformed packet and be
+                    # logged as one.
+                    fire_new_cycle = False
+                    try:
                         t = msg.get("t")
                         if t == "hello_ack":
                             self._leader_ip = sender_ip
@@ -501,7 +546,17 @@ class DisplaySyncManager:
                             self.write_status_file()
                         elif t == "sx":
                             # Vegas scroll-position sync — tiny message, renders locally
-                            self._latest_scroll_x = float(msg["x"])
+                            scroll_x = float(msg["x"])
+                            if not math.isfinite(scroll_x):
+                                # json.loads accepts the NaN/Infinity literals,
+                                # and float("nan") accepts the strings, so a
+                                # non-finite x reaches here intact. Left alone
+                                # it poisons every offset computed from it —
+                                # NaN comparisons are all false, so the
+                                # follower renders a frame it can never scroll
+                                # back from. Treat it as malformed.
+                                raise ValueError(f"non-finite scroll x: {msg['x']!r}")
+                            self._latest_scroll_x = scroll_x
                             self._last_leader_frame_time = time.time()
                             self._leader_ip = sender_ip
                             if self._follower_state == FollowerState.STANDALONE:
@@ -511,19 +566,22 @@ class DisplaySyncManager:
                                     sender_ip,
                                 )
                                 self.write_status_file()
-                                if self._on_new_cycle:
+                                fire_new_cycle = True  # build initial scroll image
-                                    self._on_new_cycle()  # build initial scroll image
                         elif t == "nc":
                             # Leader started a new scroll cycle — rebuild local image
-                            if self._on_new_cycle:
+                            fire_new_cycle = True
+                    except (KeyError, AttributeError, TypeError, ValueError) as exc:
+                        self.logger.debug("Sync: malformed control message: %s", exc)
+                        continue
+
+                    if fire_new_cycle and self._on_new_cycle:
                         self._on_new_cycle()
-                    except (json.JSONDecodeError, UnicodeDecodeError, KeyError):
-                        pass
 
             except socket.timeout:
                 continue
             except Exception as exc:
                 self.logger.debug("Sync follower recv error: %s", exc)
+                time.sleep(0.1)
 
     def _follower_announce_loop(self) -> None:
         hw = self._hw_config

src/config_manager.py

@@ -269,20 +269,47 @@ class ConfigManager:
             self.logger.error(error_msg, exc_info=True)
             raise ConfigError(error_msg, config_path=self.config_path) from e
 
+    @staticmethod
+    def _is_parallel_secrets_list(value: Any) -> bool:
+        """True for the parallel-placeholder list shape emitted by
+        ``secret_helpers.separate_secrets`` for array-item secrets: a
+        non-empty list whose elements are ALL dicts (``{}`` marks an item
+        with no secrets). Any other list-shaped secrets value is a
+        whole-key secret (e.g. a list of secret scalars)."""
+        return (isinstance(value, list) and bool(value)
+                and all(isinstance(item, dict) for item in value))
+
     def _strip_secrets_recursive(self, data_to_filter: Dict[str, Any], secrets: Dict[str, Any]) -> Dict[str, Any]:
         """Recursively remove secret keys from a dictionary."""
         result = {}
         for key, value in data_to_filter.items():
-            if key in secrets:
+            if key not in secrets:
-                if isinstance(value, dict) and isinstance(secrets[key], dict):
-                    # This key is a shared group, recurse
-                    stripped_sub_dict = self._strip_secrets_recursive(value, secrets[key])
-                    if stripped_sub_dict: # Only add if there's non-secret data left
-                        result[key] = stripped_sub_dict
-                # Else, it's a secret key at this level, so we skip it
-            else:
                 # This key is not in secrets, so we keep it
                 result[key] = value
+                continue
+            sec = secrets[key]
+            if isinstance(value, dict) and isinstance(sec, dict):
+                # This key is a shared group, recurse
+                stripped_sub_dict = self._strip_secrets_recursive(value, sec)
+                if stripped_sub_dict: # Only add if there's non-secret data left
+                    result[key] = stripped_sub_dict
+            elif isinstance(value, list) and self._is_parallel_secrets_list(sec):
+                # Parallel-list shape from separate_secrets: sec[i] holds the
+                # secret fields of value[i] ({} = item i has none). Strip each
+                # item and ALWAYS keep the list — indices must survive so the
+                # merge-on-load can realign secrets with their items. The
+                # regular list's length is authoritative: extra secrets
+                # entries are ignored.
+                stripped_items = []
+                for i, item in enumerate(value):
+                    s_item = sec[i] if i < len(sec) else {}
+                    if isinstance(item, dict) and s_item:
+                        stripped_items.append(self._strip_secrets_recursive(item, s_item))
+                    else:
+                        stripped_items.append(item)
+                result[key] = stripped_items
+            # Else: whole-key secret (scalar, list of secret scalars, or a
+            # shape mismatch) -> drop the key entirely. Never leak.
         return result
 
     def _load_secrets_for_save(self) -> Dict[str, Any]:
@@ -358,11 +385,39 @@ class ConfigManager:
             return None
 
     def _deep_merge(self, target: Dict[str, Any], source: Dict[str, Any]) -> None:
-        """Deep merge source dict into target dict."""
+        """Deep merge source dict into target dict.
+
+        Sole call site: merging config_secrets.json into the loaded config.
+        Understands the parallel-list shape separate_secrets emits for
+        array-item secrets (see _is_parallel_secrets_list): each secrets
+        list item is merged into the config list item at the same index
+        ({} placeholders skipped). The config list's length is
+        authoritative — a user deleting an array item from config.json
+        must not have it resurrected from a stale secrets entry."""
         for key, value in source.items():
             if key in target and isinstance(target[key], dict) and isinstance(value, dict):
                 self._deep_merge(target[key], value)
+            elif (key in target and isinstance(target[key], list)
+                    and self._is_parallel_secrets_list(value)):
+                tlist = target[key]
+                for i, s_item in enumerate(value):
+                    if i >= len(tlist):
+                        # Interpolate only config-side data here — nothing
+                        # iterated out of the secrets dict (not even the key
+                        # name) may reach the log.
+                        self.logger.warning(
+                            "A secrets list is longer than the config list it "
+                            "parallels (config has %d item(s)); ignoring the "
+                            "extra entries", len(tlist))
+                        break
+                    if not s_item:
+                        continue  # {} placeholder: item i has no secrets
+                    if isinstance(tlist[i], dict):
+                        self._deep_merge(tlist[i], s_item)
                     else:
+                        tlist[i] = s_item  # shape drift; the secret wins
+            else:
+                # Scalars AND whole-secret scalar arrays: replace (legacy).
                 target[key] = value
 
     def _create_config_from_template(self) -> None:

src/display_controller.py

@@ -44,6 +44,20 @@ from src.common.sync_manager import DisplaySyncManager, SyncRole
 # Get logger with consistent configuration
 logger = get_logger(__name__)
 
+# How long startup will wait for plugins to fetch their first data before
+# showing anything. Each plugin's update blocks for up to the executor's 30s
+# timeout and they run one after another, so the uncapped total is the sum of
+# every slow plugin: 82 seconds on the worst boot measured, with a blank panel
+# throughout. Whatever does not finish in time is picked up by the scheduled
+# update tick moments later, with the display already running.
+_INITIAL_UPDATE_BUDGET_SECONDS = 20.0
+
+# The least budget worth starting a plugin with. Below this the plugin is
+# deferred instead: granting it a floor would let the pass run past its
+# deadline, and granting it the true remainder would record a timeout for a
+# slot it never had a chance to use.
+_MIN_INITIAL_UPDATE_TIMEOUT_SECONDS = 2.0
+
 # Vegas mode import (lazy loaded to avoid circular imports)
 _vegas_mode_imported = False
 VegasModeCoordinator = None
@@ -90,7 +104,8 @@ class DisplayController:
         # Validate startup configuration
         try:
             from src.startup_validator import StartupValidator
-            validator = StartupValidator(self.config_manager)
+            validator = StartupValidator(self.config_manager,
+                                         cache_manager=self.cache_manager)
             is_valid, errors, warnings = validator.validate_all()
             
             if warnings:
@@ -166,6 +181,16 @@ class DisplayController:
         self.plugin_modes = {}  # mode -> plugin_instance mapping for plugin-first dispatch
         self.mode_to_plugin_id: Dict[str, str] = {}
         self.plugin_display_modes: Dict[str, List[str]] = {}
+        # plugin_display_modes is mutated only by _register_loaded_plugin /
+        # _unregister_plugin on the render thread, but the config-watcher
+        # thread reads it in _enabled_plugin_not_running. Both mutation sites
+        # run during reconcile (rare), so this lock never touches the per-frame
+        # path -- the hot-path reads are same-thread as the writes.
+        self._plugin_modes_lock = threading.Lock()
+        # Guards the consume-and-clear of _pending_plugin_reconcile. Only taken
+        # when a reconcile is actually pending or a config change arrives, both
+        # rare -- the per-frame path just reads the bool.
+        self._reconcile_flag_lock = threading.Lock()
         # Per-plugin config-change callbacks, kept so we can unsubscribe a
         # plugin when it is disabled live.
         self._plugin_config_callbacks: Dict[str, Callable] = {}
@@ -258,7 +283,8 @@ class DisplayController:
             # Validate plugins after plugin manager is created
             try:
                 from src.startup_validator import StartupValidator
-                validator = StartupValidator(self.config_manager, self.plugin_manager)
+                validator = StartupValidator(self.config_manager, self.plugin_manager,
+                                             cache_manager=self.cache_manager)
                 is_valid, errors, warnings = validator.validate_all()
                 
                 if warnings:
@@ -447,7 +473,9 @@ class DisplayController:
             self._refresh_config_cache(new_config)
             # If a plugin was enabled/disabled, flag a reconcile for the main
             # loop to apply (loading/unloading off the watcher thread is unsafe).
-            if self._enabled_set_changed(old_config, new_config):
+            if (self._enabled_set_changed(old_config, new_config)
+                    or self._enabled_plugin_not_running(new_config)):
+                with self._reconcile_flag_lock:
                     self._pending_plugin_reconcile = True
 
         self.config_service.subscribe(_controller_config_change)
@@ -461,7 +489,7 @@ class DisplayController:
         # Initial data update for plugins (ensures data available on first display)
         logger.info("Performing initial plugin data update...")
         update_start = time.time()
-        self._update_modules()
+        self._update_modules(deadline=update_start + _INITIAL_UPDATE_BUDGET_SECONDS)
         logger.info("Initial plugin update completed in %.3f seconds", time.time() - update_start)
 
         # Initialize Vegas mode coordinator
@@ -817,14 +845,42 @@ class DisplayController:
             self._cached_target_brightness = normal_brightness  # persist for minute-gate
             return normal_brightness
 
-    def _update_modules(self):
+    def _update_modules(self, deadline: Optional[float] = None):
-        """Update all plugin modules."""
+        """Update all plugin modules.
+
+        Args:
+            deadline: Wall-clock time after which remaining plugins are left
+                for the scheduled update tick instead of being waited on. Each
+                update blocks this thread for up to the executor's timeout, and
+                they run one after another, so without a bound the total is the
+                sum of every slow plugin on the system. Measured at startup on
+                a live rig: 82 seconds, 55 and 26 on the two boots before -- all
+                of it with nothing on the panel.
+        """
         if not self.plugin_manager:
             return
 
         # Update all loaded plugins
         plugins_dict = getattr(self.plugin_manager, 'loaded_plugins', None) or getattr(self.plugin_manager, 'plugins', {})
+        deferred = []
         for plugin_id, plugin_instance in plugins_dict.items():
+            update_timeout = None
+            if deadline is not None:
+                update_timeout = deadline - time.time()
+                if update_timeout < _MIN_INITIAL_UPDATE_TIMEOUT_SECONDS:
+                    # Too little left to be worth starting. Deferring rather
+                    # than granting a floor keeps the budget a real ceiling --
+                    # clamping up to a minimum let a plugin that began with a
+                    # sliver left run on past the deadline -- and a plugin
+                    # handed a slot it cannot use would just be recorded as
+                    # having timed out.
+                    #
+                    # Nothing is lost either way: a plugin that has never
+                    # updated is immediately due, so run_scheduled_updates()
+                    # picks it up within seconds, with the display already
+                    # running.
+                    deferred.append(plugin_id)
+                    continue
             # Check circuit breaker before attempting update
             if hasattr(self.plugin_manager, 'health_tracker') and self.plugin_manager.health_tracker:
                 if self.plugin_manager.health_tracker.should_skip_plugin(plugin_id):
@@ -833,7 +889,13 @@ class DisplayController:
             
             # Use PluginExecutor if available for safe execution
             if hasattr(self.plugin_manager, 'plugin_executor'):
-                success = self.plugin_manager.plugin_executor.execute_update(plugin_instance, plugin_id)
+                # The remaining budget is the timeout, so the pass cannot
+                # run past its deadline. Bounding the loop alone did not do
+                # it: the last plugin to start could still block for the
+                # executor's full 30s, which turned a 20s budget into a 31.8s
+                # pass on the rig.
+                success = self.plugin_manager.plugin_executor.execute_update(
+                    plugin_instance, plugin_id, timeout=update_timeout)
                 if success and hasattr(self.plugin_manager, 'plugin_last_update'):
                     self.plugin_manager.plugin_last_update[plugin_id] = time.time()
             else:
@@ -852,6 +914,12 @@ class DisplayController:
                     if hasattr(self.plugin_manager, 'health_tracker') and self.plugin_manager.health_tracker:
                         self.plugin_manager.health_tracker.record_failure(plugin_id, exc)
 
+        if deferred:
+            logger.info(
+                "Initial update budget spent; %d plugin(s) left to the update "
+                "tick so the display can start: %s",
+                len(deferred), ", ".join(deferred))
+
     def _tick_plugin_updates_for_vegas(self) -> None:
         """Run scheduled plugin updates and tell Vegas mode which plugins
         actually got fresh data, so it can hot-swap them into the scroll
@@ -1638,6 +1706,12 @@ class DisplayController:
                 logger.warning("Error checking live priority for %s: %s", mode_name, e)
         return live
 
+    def _vegas_keeps_live_in_ticker(self) -> bool:
+        """Whether live content should stay in the ticker instead of preempting it."""
+        coordinator = getattr(self, 'vegas_coordinator', None)
+        config = getattr(coordinator, 'vegas_config', None)
+        return bool(getattr(config, 'live_in_ticker', False))
+
     def _check_live_priority(self, advance=False):
         """Return the live-priority mode to display, or None if nothing is live.
 
@@ -1687,11 +1761,12 @@ class DisplayController:
                 # rebuilding available_modes happens here on the render thread so
                 # it can't race with rendering. Deferred while on-demand is active
                 # (the flag stays set) so we don't fight its temporary-enable.
+                # The lock-free read is a fast path only; it can be a false
+                # negative (the watcher setting the flag just after it is read
+                # is seen next iteration), never a false positive that loses a
+                # request.
                 if self._pending_plugin_reconcile and not self.on_demand_active:
-                    # Only clear the flag on success -- a retryable failure
+                    self._service_pending_reconcile()
-                    # (e.g. discovery) leaves it set so the request isn't lost.
-                    if self._reconcile_enabled_plugins():
-                        self._pending_plugin_reconcile = False
 
                 if not self.available_modes:
                     # Nothing to render yet. Re-check _pending_plugin_reconcile
@@ -1851,14 +1926,24 @@ class DisplayController:
                 # Check for live priority content and switch to it immediately.
                 # advance=True so multiple simultaneously-live games take turns
                 # (round-robin) instead of pinning to the first plugin.
-                if not self.on_demand_active and not wifi_status_data:
+                # Skipped when the ticker is keeping live content: switching
+                # the rotation underneath Vegas would move current_mode_index
+                # and stash a resume point for a takeover that never happens.
+                if (not self.on_demand_active and not wifi_status_data
+                        and not (self._is_vegas_mode_active()
+                                 and self._vegas_keeps_live_in_ticker())):
                     live_priority_mode = self._check_live_priority(advance=True)
                     self._apply_live_priority(live_priority_mode)
 
                 # Vegas scroll mode - continuous ticker across all plugins
                 # Priority: on-demand > wifi-status > live-priority > vegas > normal rotation
                 if self._is_vegas_mode_active() and not wifi_status_data:
-                    live_mode = self._check_live_priority()
+                    # Live content normally preempts the ticker entirely. With
+                    # vegas_scroll.live_in_ticker the marquee keeps running and
+                    # the live plugin takes extra turns inside it instead --
+                    # see StreamManager._apply_priority_weights.
+                    live_mode = (None if self._vegas_keeps_live_in_ticker()
+                                 else self._check_live_priority())
                     if not live_mode:
                         try:
                             # Run Vegas mode iteration
@@ -2741,6 +2826,7 @@ class DisplayController:
             logger.debug("Using manifest display_modes for %s: %s", plugin_id, display_modes)
         if not (isinstance(display_modes, list) and display_modes):
             display_modes = [plugin_id]
+        with self._plugin_modes_lock:
             self.plugin_display_modes[plugin_id] = list(display_modes)
 
         # Subscribe to config changes for per-plugin hot-reload. Bind plugin_id
@@ -2775,6 +2861,7 @@ class DisplayController:
     def _unregister_plugin(self, plugin_id: str) -> None:
         """Remove a plugin's modes, config subscription and instance, then
         unload it. Used by live disable hot-reload."""
+        with self._plugin_modes_lock:
             modes = self.plugin_display_modes.pop(plugin_id, [])
         for mode in modes:
             if mode in self.available_modes:
@@ -2820,6 +2907,67 @@ class DisplayController:
             }
         return enabled_map(old_config) != enabled_map(new_config)
 
+    def _service_pending_reconcile(self) -> None:
+        """Consume a pending reconcile request and run it.
+
+        The request is consumed BEFORE reconciling, not cleared after. Clearing
+        after would drop any config change that lands while reconcile is
+        running: reconcile has already read its config by then, so the clear
+        erases a request it never served and the newest config never
+        reconciles -- the same "your save did nothing" failure this whole path
+        exists to prevent. Consuming first means such a request stays set and
+        is picked up on the next pass.
+
+        A retryable failure (e.g. discovery) re-arms the flag.
+        """
+        with self._reconcile_flag_lock:
+            pending = self._pending_plugin_reconcile
+            self._pending_plugin_reconcile = False
+        if pending and not self._reconcile_enabled_plugins():
+            with self._reconcile_flag_lock:
+                self._pending_plugin_reconcile = True
+
+    def _enabled_plugin_not_running(self, new_config: Dict[str, Any]) -> bool:
+        """True when a discovered plugin is enabled in config but not running.
+
+        ``_enabled_set_changed`` compares only top-level ``enabled`` flags, which
+        misses the case that strands a plugin: one whose ``validate_config()``
+        returned False is absent from the running set, and the edit that fixes it
+        (enabling a league, filling in an API key) lives *nested* inside that
+        plugin's own section. No top-level flag changes, so no reconcile is
+        queued, and the save that should have fixed it appears to do nothing --
+        only toggling some unrelated plugin recovers it. hockey-scoreboard sat
+        enabled-but-absent on a live rig for four days this way.
+
+        Deliberately narrow: it fires only for ids the plugin manager has
+        actually discovered, so non-plugin sections that carry their own
+        ``enabled`` flag (``schedule``, ``display``, ...) don't queue a reconcile
+        on every save. In the steady state -- everything enabled is loaded --
+        this is False and costs nothing. That matters because reconcile runs
+        ``discover_plugins()`` on the render thread, where a needless
+        filesystem scan per config save would show up as a frame hitch.
+
+        Runs on the config-watcher thread, so both mappings it reads are
+        snapshotted under the lock that guards their writes.
+        """
+        if self.plugin_manager is None:
+            return False
+        # Two snapshots, each taken under its own lock and never nested, so a
+        # half-written mapping is never observed and this can't deadlock
+        # against discovery (which holds the discovery lock while rebuilding).
+        try:
+            known = self.plugin_manager.discovered_plugin_ids()
+        except AttributeError:
+            # Older manager without the accessor: fall back to a plain read.
+            known = set(getattr(self.plugin_manager, 'plugin_manifests', ()) or ())
+        with self._plugin_modes_lock:
+            running = set(self.plugin_display_modes)
+        for key, value in new_config.items():
+            if (key in known and isinstance(value, dict)
+                    and value.get('enabled', False) and key not in running):
+                return True
+        return False
+
     def _reconcile_enabled_plugins(self) -> bool:
         """Load/unload plugins so the running set matches the enabled set in
         config. Runs on the main display thread (never the config-watcher

src/display_manager.py

@@ -25,6 +25,7 @@ the same object.
 
 import json
 import os
+import socket
 import tempfile
 if os.getenv("EMULATOR", "false") == "true":
     from RGBMatrixEmulator import RGBMatrix, RGBMatrixOptions
@@ -258,6 +259,26 @@ class DisplayManager:
         # Initialize managers
         # Calendar manager is now initialized by DisplayController
         
+    # Orientation setting -> rpi-rgb-led-matrix "Rotate:<deg>" pixel-mapper suffix.
+    # "normal" needs no suffix since 0 degrees is the identity transform.
+    _ORIENTATION_ROTATE_DEGREES = {'normal': None, '90': 90, '180': 180, '270': 270}
+
+    def _build_pixel_mapper_config(self, hardware_config: dict) -> str:
+        """Compose the raw pixel_mapper_config string with the orientation setting.
+
+        `pixel_mapper_config` stays available as a free-form advanced field (e.g.
+        for "U-mapper" chain layouts); `orientation` is the user-facing dropdown
+        for physical mounting (e.g. panels mounted upside down) and is appended as
+        a "Rotate:<deg>" mapper rather than overwriting any existing config.
+        """
+        base_mapper = (hardware_config.get('pixel_mapper_config') or '').strip()
+        orientation = hardware_config.get('orientation', 'normal')
+        degrees = self._ORIENTATION_ROTATE_DEGREES.get(orientation)
+        if degrees is None:
+            return base_mapper
+        rotate_mapper = f'Rotate:{degrees}'
+        return f'{base_mapper};{rotate_mapper}' if base_mapper else rotate_mapper
+
     def _setup_matrix(self):
         """Initialize the RGB matrix with configuration settings."""
         _init_error_str = None
@@ -283,7 +304,7 @@ class DisplayManager:
             options.pwm_bits = hardware_config.get('pwm_bits', 10)
             options.pwm_lsb_nanoseconds = hardware_config.get('pwm_lsb_nanoseconds', 150)
             options.led_rgb_sequence = hardware_config.get('led_rgb_sequence', 'RGB')
-            options.pixel_mapper_config = hardware_config.get('pixel_mapper_config', '')
+            options.pixel_mapper_config = self._build_pixel_mapper_config(hardware_config)
             options.row_address_type = hardware_config.get('row_address_type', 0)
             options.multiplexing = hardware_config.get('multiplexing', 0)
             options.panel_type = hardware_config.get('panel_type', '')
@@ -343,6 +364,7 @@ class DisplayManager:
             # Create image with the (logical) display dimensions
             self.image = Image.new('RGB', (self.matrix.width, self.matrix.height))
             self.draw = ImageDraw.Draw(self.image)
+            self.draw.fontmode = "1"  # 1-bit text: the panel has no partial brightness, so AA only smears glyphs.
             logger.info(f"Image canvas created with dimensions: {self.matrix.width}x{self.matrix.height}")
             
             # Initialize font with Press Start 2P
@@ -382,6 +404,7 @@ class DisplayManager:
 
             self.image = Image.new('RGB', (fallback_width, fallback_height))
             self.draw = ImageDraw.Draw(self.image)
+            self.draw.fontmode = "1"  # 1-bit text: the panel has no partial brightness, so AA only smears glyphs.
             # Simple fallback visualization so web UI shows a realistic canvas
             try:
                 self.draw.rectangle([0, 0, fallback_width - 1, fallback_height - 1], outline=(255, 0, 0))
@@ -497,6 +520,91 @@ class DisplayManager:
             logger.warning(f"[BRIGHTNESS] Matrix does not support brightness property: {e}", exc_info=True)
             return -1
 
+    @staticmethod
+    def _local_ip() -> Optional[str]:
+        """This device's address on the network it routes through, or None.
+
+        Deliberately not `hostname -I` or a systemctl probe for AP mode, which
+        is how the web launcher does it: both spawn processes with multi-second
+        timeouts, and this runs on the startup path the rest of this change
+        exists to shorten. Connecting a UDP socket sends no packets -- it only
+        asks the kernel which source address it would use -- so it costs
+        microseconds and works with the network down, as long as a route
+        exists.
+        """
+        sock = None
+        try:
+            sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+            sock.settimeout(0.2)
+            sock.connect(("8.8.8.8", 80))  # nosec B104 - no traffic; selects a route
+            ip = sock.getsockname()[0]
+            return ip if ip and not ip.startswith("127.") else None
+        except OSError:
+            return None
+        finally:
+            if sock is not None:
+                try:
+                    sock.close()
+                except OSError:
+                    pass
+
+    def _fitting_font(self, lines, width):
+        """The largest font from the usual ladder that fits every line."""
+        candidates = [self.font,
+                      ("assets/fonts/4x6-font.ttf", 6)]
+        for candidate in candidates:
+            try:
+                font = candidate
+                if isinstance(candidate, tuple):
+                    font = ImageFont.truetype(candidate[0], candidate[1])
+                if all(self.draw.textlength(t, font=font) <= width for t in lines):
+                    return font
+            except (OSError, ValueError, AttributeError):
+                continue
+        return self.font
+
+    def _draw_startup_banner(self, lines, width: int, height: int) -> None:
+        """Centre `lines` over whatever the test pattern already drew.
+
+        This screen stays on the panel for the whole initial plugin update, and
+        on a headless Pi it is the only place the device's address appears
+        without going looking for it -- so it has to be readable off a wall,
+        not merely present.
+
+        The font is chosen to fit rather than fixed at 8px: "Initializing" is
+        96px in PressStart2P, which ran off the side of a 64px panel even
+        before an address was added. And the pattern is punched out behind the
+        text, because the diagonal runs through the middle of the panel, which
+        is exactly where this sits.
+
+        The text stays blue. It is not decoration: the pattern draws one pure
+        channel per element -- red border, green diagonal, blue text -- so that
+        a glance at the panel says whether led_rgb_sequence is right. Swap the
+        wiring to BGR and the border comes up blue and this text red. Drawing
+        it white would light all three channels and destroy the only blue
+        reference on the screen, which is why it is worth a comment rather
+        than a quiet preference.
+        """
+        if not lines:
+            return
+        font = self._fitting_font(lines, width - 2)
+        line_height = self.draw.textbbox((0, 0), "Ag", font=font)[3] + 1
+        block_height = line_height * len(lines)
+        block_top = max(1, (height - block_height) // 2)
+        block_width = max(self.draw.textlength(t, font=font) for t in lines)
+        block_left = max(0, (width - block_width) // 2)
+
+        self.draw.rectangle(
+            [block_left - 2, block_top - 1,
+             block_left + block_width + 1, block_top + block_height],
+            fill=(0, 0, 0))
+
+        for row, line in enumerate(lines):
+            line_width = self.draw.textlength(line, font=font)
+            self.draw.text(
+                (max(0, (width - line_width) // 2), block_top + row * line_height),
+                line, font=font, fill=(0, 0, 255))
+
     def _draw_test_pattern(self):
         """Draw a test pattern to verify the display is working."""
         try:
@@ -516,8 +624,11 @@ class DisplayManager:
             # Draw a diagonal line
             self.draw.line([0, 0, self.matrix.width-1, self.matrix.height-1], fill=(0, 255, 0))
             
-            # Draw some text - changed from "TEST" to "Initializing" with smaller font
+            lines = ["Initializing"]
-            self.draw.text((10, 10), "Initializing", font=self.font, fill=(0, 0, 255))
+            ip = self._local_ip()
+            if ip:
+                lines.append(ip)
+            self._draw_startup_banner(lines, self.matrix.width, self.matrix.height)
             
             # Update the display once after everything is drawn
             self.update_display()
@@ -596,6 +707,7 @@ class DisplayManager:
             # self.image, so swapping the buffer below is enough on its own.
             self.image = Image.new('RGB', (target_w, target_h))
             self.draw = ImageDraw.Draw(self.image)
+            self.draw.fontmode = "1"  # 1-bit text: the panel has no partial brightness, so AA only smears glyphs.
             yield
         finally:
             self.matrix = real_matrix
@@ -705,6 +817,7 @@ class DisplayManager:
                 
                 self.image = Image.new('RGB', (width, height))
                 self.draw = ImageDraw.Draw(self.image)
+                self.draw.fontmode = "1"  # 1-bit text: the panel has no partial brightness, so AA only smears glyphs.
                 logger.debug("Cleared display in fallback mode")
                 return
                 
@@ -716,6 +829,7 @@ class DisplayManager:
             # Create a new black image
             self.image = Image.new('RGB', (self.matrix.width, self.matrix.height))
             self.draw = ImageDraw.Draw(self.image)
+            self.draw.fontmode = "1"  # 1-bit text: the panel has no partial brightness, so AA only smears glyphs.
 
             if not self._capture_mode_active:
                 # Clear both canvases and the underlying matrix to ensure no artifacts.
@@ -1149,6 +1263,7 @@ class DisplayManager:
             try:
                 self.image = Image.new('RGB', (self.width, self.height))
                 self.draw = ImageDraw.Draw(self.image)
+                self.draw.fontmode = "1"  # 1-bit text: the panel has no partial brightness, so AA only smears glyphs.
             except (OSError, RuntimeError, ValueError, MemoryError):
                 logger.debug("Canvas reset during cleanup failed", exc_info=True)
         # Reset the singleton state when cleaning up

src/dynamic_team_resolver.py

@@ -168,9 +168,13 @@ class DynamicTeamResolver:
                 # Sort by ranking (1, 2, 3, etc.)
                 sorted_rankings = dict(sorted(rankings.items(), key=lambda x: x[1]))
 
-                # Cache the results
+                # Cache the results ON THE CLASS. Assigning through self
-                self._rankings_cache = sorted_rankings
+                # would create instance attributes that shadow the shared
-                self._cache_timestamp = current_time
+                # class-level cache, making it per-instance — and every
+                # scoreboard constructs its own resolver, so the cache
+                # would never actually be shared.
+                DynamicTeamResolver._rankings_cache = sorted_rankings
+                DynamicTeamResolver._cache_timestamp = current_time
                 
                 self.logger.info(f"Fetched rankings for {len(sorted_rankings)} teams")
                 return sorted_rankings
@@ -216,9 +220,11 @@ class DynamicTeamResolver:
         return any(pattern in team_name.upper() for pattern in dynamic_patterns)
     
     def clear_cache(self):
-        """Clear the rankings cache to force fresh data on next request."""
+        """Clear the SHARED rankings cache to force fresh data on next
-        self._rankings_cache = {}
+        request. Writes through the class — assigning via self would only
-        self._cache_timestamp = 0
+        shadow the shared cache for this instance."""
+        DynamicTeamResolver._rankings_cache = {}
+        DynamicTeamResolver._cache_timestamp = 0
         self.logger.info("Cleared dynamic team rankings cache")
 
 

src/logging_config.py

@@ -5,6 +5,7 @@ Provides consistent logging configuration across the LEDMatrix application.
 Supports structured logging with context information and appropriate log levels.
 """
 
+import copy
 import logging
 import sys
 import os
@@ -65,8 +66,12 @@ class ContextualFormatter(logging.Formatter):
         self.include_context = include_context
     
     def format(self, record: logging.LogRecord) -> str:
-        """Format log record with context."""
+        """Format log record with context.
-        # Add context to message if present
+
+        Works on a shallow copy of the record: a record is formatted once
+        PER HANDLER, so mutating record.msg in place (the old behavior)
+        prepended the context prefix again for every additional handler.
+        """
         if self.include_context:
             context_parts = []
 
@@ -81,6 +86,7 @@ class ContextualFormatter(logging.Formatter):
                     context_parts.append(f"[{key}: {value}]")
 
             if context_parts:
+                record = copy.copy(record)
                 record.msg = ' '.join(context_parts) + ' ' + str(record.msg)
 
         return super().format(record)
@@ -124,7 +130,12 @@ def setup_logging(
     # Console handler (always add)
     console_handler = logging.StreamHandler(sys.stdout)
     console_handler.setLevel(level)
-    console_handler.setFormatter(formatter)
+    # Under systemd, tag each line so the journal records the real severity
+    # rather than filing everything as informational. The file handler below
+    # keeps the plain formatter: the prefix is meaningful to journald and noise
+    # anywhere else.
+    console_handler.setFormatter(
+        JournalPriorityFormatter(formatter) if _under_systemd() else formatter)
     root_logger.addHandler(console_handler)
     
     # File handler (if specified)
@@ -139,6 +150,84 @@ def setup_logging(
             sys.stderr.write(f"Warning: Could not set up file logging to {log_file}: {e}\n")
 
 
+#: syslog priorities, which is what systemd parses from a "<N>" prefix on
+#: stdout. Mapped from Python's levels.
+_SYSLOG_PRIORITY = {
+    logging.CRITICAL: 2,   # LOG_CRIT
+    logging.ERROR: 3,      # LOG_ERR
+    logging.WARNING: 4,    # LOG_WARNING
+    logging.INFO: 6,       # LOG_INFO
+    logging.DEBUG: 7,      # LOG_DEBUG
+}
+
+
+class JournalPriorityFormatter(logging.Formatter):
+    """Wraps a formatter, prefixing each line with its syslog priority.
+
+    Under systemd everything this process writes to stdout lands in the journal
+    as PRIORITY=6, whatever the Python level was. Measured on a live rig: 55
+    ERROR lines and 13 WARNING lines in a day, every one of them recorded as
+    informational, so `journalctl -p err -u ledmatrix` returned nothing at all
+    while errors were being logged. Anyone triaging has to grep the message
+    text instead, which is both slower and wrong -- a search for "oom" matches
+    the radar logging "zoom=9".
+
+    systemd reads a leading "<N>" on each line and uses it as the priority
+    (sd-daemon(3)), so this needs no extra dependency. Multi-line records get
+    the prefix on every line, since the journal splits them and an unprefixed
+    continuation would fall back to the default.
+    """
+
+    def __init__(self, inner: logging.Formatter):
+        super().__init__()
+        self._inner = inner
+
+    @property
+    def inner(self) -> logging.Formatter:
+        """The formatter doing the actual work.
+
+        Whether journald tagging is applied depends on JOURNAL_STREAM, so it is
+        on under systemd and off in a terminal -- and anything asserting which
+        formatter setup_logging() selected would otherwise get a different
+        answer in CI than on a developer's machine. Exposing the inner one lets
+        those checks stay about format_type, which is what they mean.
+        """
+        return self._inner
+
+    def format(self, record: logging.LogRecord) -> str:
+        text = self._inner.format(record)
+        prefix = f"<{_SYSLOG_PRIORITY.get(record.levelno, 6)}>"
+        return "\n".join(prefix + line for line in text.split("\n"))
+
+
+def _under_systemd() -> bool:
+    """True when stdout really is the journal.
+
+    systemd sets JOURNAL_STREAM to "dev:ino" for services whose output it
+    captures. Presence alone is not enough to act on: the variable is
+    inherited by child processes and survives redirection, so a subprocess
+    whose stdout is a pipe or a file still sees it and would emit the "<N>"
+    priority prefixes as literal noise into that output. systemd's own
+    guidance is to fstat the descriptor and compare st_dev/st_ino, which is
+    what distinguishes "the journal is somewhere in my ancestry" from "my
+    stdout is the journal".
+    """
+    declared = os.environ.get("JOURNAL_STREAM")
+    if not declared:
+        return False
+    try:
+        dev_text, ino_text = declared.split(":", 1)
+        declared_ids = (int(dev_text), int(ino_text))
+    except (ValueError, AttributeError):
+        return False
+    try:
+        stat_result = os.fstat(sys.stdout.fileno())
+    except (OSError, ValueError, AttributeError):
+        # No usable stdout: captured by pytest, detached, or already closed.
+        return False
+    return (stat_result.st_dev, stat_result.st_ino) == declared_ids
+
+
 class PluginLoggerAdapter(logging.LoggerAdapter):
     """LoggerAdapter that stamps every record with its plugin_id.
 
@@ -224,8 +313,11 @@ def log_warning(logger: logging.Logger, message: str, **kwargs) -> None:
 
 
 def log_error(logger: logging.Logger, message: str, **kwargs) -> None:
-    """Log error message with context."""
+    """Log error message with context. Defaults exc_info=True; a caller
-    log_with_context(logger, logging.ERROR, message, **kwargs, exc_info=True)
+    passing exc_info explicitly wins (the old hardcoded keyword raised
+    TypeError on that duplicate)."""
+    kwargs.setdefault('exc_info', True)
+    log_with_context(logger, logging.ERROR, message, **kwargs)
 
 
 def log_debug(logger: logging.Logger, message: str, **kwargs) -> None:

src/logo_downloader.py

@@ -14,6 +14,7 @@ import json
 from typing import Dict, List, Optional, Tuple
 from pathlib import Path
 from PIL import Image, ImageDraw, ImageFont
+from PIL.PngImagePlugin import PngInfo
 from requests.adapters import HTTPAdapter
 from urllib3.util.retry import Retry
 from src.common.permission_utils import (
@@ -25,6 +26,96 @@ from src.common.permission_utils import (
 
 logger = logging.getLogger(__name__)
 
+#: PNG text key stamped into a generated placeholder so a later run can tell it
+#: apart from a real logo that happens to be small.
+PLACEHOLDER_MARKER = "ledmatrix_placeholder"
+
+#: Geometry of a generated placeholder, used to recognise ones written before
+#: the marker existed. Those are already on users' disks and would otherwise
+#: never be retried.
+PLACEHOLDER_SIZE = (64, 64)
+PLACEHOLDER_BG = (100, 100, 100, 255)
+
+#: How long a placeholder is trusted before the real logo is attempted again.
+#: A placeholder means the download failed, and download failures are usually
+#: transient (no network at boot, ESPN blipping). Retrying every frame would
+#: hammer the API from a Pi that is also driving a panel; never retrying leaves
+#: the team a grey box forever, which is the bug this exists to avoid.
+PLACEHOLDER_RETRY_SECONDS = 6 * 60 * 60
+
+
+def is_placeholder_logo(filepath: Path) -> bool:
+    """True if the file at ``filepath`` is a generated placeholder, not a logo.
+
+    Checks the marker first, then falls back to matching the placeholder's
+    exact geometry and background colour so files written before the marker was
+    introduced are still recognised.
+    """
+    try:
+        with Image.open(filepath) as img:
+            if img.info.get(PLACEHOLDER_MARKER):
+                return True
+            if img.size != PLACEHOLDER_SIZE:
+                return False
+            return img.convert("RGBA").getpixel((0, 0)) == PLACEHOLDER_BG
+    except Exception:
+        # Unreadable file: not provably a placeholder, and the caller's own
+        # error handling is better placed to deal with it.
+        return False
+
+
+def should_attempt_download(filepath: Path, force_download: bool = False) -> bool:
+    """Whether a real logo is worth (re)fetching for ``filepath``.
+
+    True when nothing is there, when the caller forced it, or when what is
+    there is a placeholder old enough to retry. A *fresh* placeholder says a
+    download just failed, so retrying it immediately would hammer the API for
+    a result that is very unlikely to have changed.
+    """
+    if force_download or not filepath.exists():
+        return True
+    if not is_placeholder_logo(filepath):
+        return False
+    age = placeholder_age_seconds(filepath)
+    return age is None or age >= PLACEHOLDER_RETRY_SECONDS
+
+
+def refresh_placeholder_timestamp(filepath: Path) -> bool:
+    """Restamp a placeholder so a failed retry restarts the back-off clock.
+
+    Without this a stale placeholder stays stale: every later call sees an
+    expired timestamp, retries, fails, and leaves the timestamp untouched --
+    which is a download attempt per call, the opposite of what the back-off is
+    for.
+    """
+    try:
+        if not is_placeholder_logo(filepath):
+            return False
+        metadata = PngInfo()
+        metadata.add_text(PLACEHOLDER_MARKER, str(time.time()))
+        with Image.open(filepath) as img:
+            img.copy().save(filepath, "PNG", pnginfo=metadata)
+        return True
+    except Exception:
+        logger.debug("Could not refresh placeholder timestamp for %s", filepath,
+                     exc_info=True)
+        return False
+
+
+def placeholder_age_seconds(filepath: Path) -> Optional[float]:
+    """Seconds since a placeholder was written, or None if unknown."""
+    try:
+        with Image.open(filepath) as img:
+            stamped = img.info.get(PLACEHOLDER_MARKER)
+        if stamped and stamped != "1":
+            return max(0.0, time.time() - float(stamped))
+    except Exception:
+        pass
+    try:
+        return max(0.0, time.time() - filepath.stat().st_mtime)
+    except OSError:
+        return None
+
 class LogoDownloader:
     """Centralized logo downloader for team logos from ESPN API."""
     
@@ -499,8 +590,10 @@ class LogoDownloader:
             filename = f"{self.normalize_abbreviation(abbreviation)}.png"
             filepath = Path(logo_dir) / filename
             
-            # Skip if already exists and not forcing download
+            # A placeholder does not count as existing -- it is a previous
-            if filepath.exists() and not force_download:
+            # failure, and a bulk pass is exactly where it should get another
+            # chance, subject to the same back-off as everywhere else.
+            if not should_attempt_download(filepath, force_download):
                 logger.debug(f"Skipping {display_name}: {filename} already exists")
                 continue
             
@@ -559,8 +652,9 @@ class LogoDownloader:
             filename = f"{self.normalize_abbreviation(abbreviation)}.png"
             filepath = Path(logo_dir) / filename
             
-            # Skip if already exists and not forcing download
+            # Same eligibility rule as every other download site: a stale
-            if filepath.exists() and not force_download:
+            # placeholder is a failed download, not a logo.
+            if not should_attempt_download(filepath, force_download):
                 logger.debug(f"Skipping {display_name} ({category}, {conference}): {filename} already exists")
                 continue
             
@@ -674,7 +768,12 @@ class LogoDownloader:
                 # Fallback without font
                 draw.text((16, 24), text, fill=(255, 255, 255, 255))
             
-            logo.save(filepath)
+            # Stamp it so a later run can tell this apart from a real logo and
+            # retry the download, instead of treating the file's existence as
+            # proof the logo was fetched.
+            metadata = PngInfo()
+            metadata.add_text(PLACEHOLDER_MARKER, str(time.time()))
+            logo.save(filepath, "PNG", pnginfo=metadata)
 
             # Set proper file permissions after saving
             ensure_file_permissions(filepath, get_assets_file_mode())
@@ -770,9 +869,18 @@ def download_missing_logo(league: str, team_id: str, team_abbreviation: str, log
     # Use the exact filepath that was passed in (respects config settings)
     filepath = logo_path
     
-    if filepath.exists():
+    if filepath.exists() and not should_attempt_download(filepath):
+        # Either a real logo, or a placeholder too fresh to be worth retrying.
         logger.debug(f"Logo already exists for {team_abbreviation} ({league})")
         return True
+    if filepath.exists():
+        # A placeholder is a *failed* download wearing the real logo's
+        # filename. Treating it as "already exists" is what pinned a team to a
+        # grey box permanently after one transient failure.
+        logger.info(
+            "Logo for %s (%s) is a placeholder from a failed download; "
+            "retrying the real logo", team_abbreviation, league,
+        )
     
     # Try to download the real logo first
     logger.info(f"Attempting to download logo for {team_abbreviation} from {league}")

src/plugin_system/base_plugin.py

@@ -364,8 +364,10 @@ class BasePlugin(ABC):
                 # Handle None case
                 if duration is None:
                     pass  # Fall through to config
-                # Try to convert to float if it's a number or numeric string
+                # Try to convert to float if it's a number or numeric string.
-                elif isinstance(duration, (int, float)):
+                # bool is excluded: it's an int subclass, and True would
+                # otherwise read as a 1-second duration.
+                elif isinstance(duration, (int, float)) and not isinstance(duration, bool):
                     if duration > 0:
                         return float(duration)
                     else:
@@ -403,8 +405,9 @@ class BasePlugin(ABC):
         # Fall back to config
         config_duration = self.config.get("display_duration", 15.0)
         try:
-            # Ensure config value is also a valid float
+            # Ensure config value is also a valid float (bool excluded — an
-            if isinstance(config_duration, (int, float)):
+            # int subclass that would otherwise read True as 1 second)
+            if isinstance(config_duration, (int, float)) and not isinstance(config_duration, bool):
                 if config_duration > 0:
                     return float(config_duration)
                 else:
@@ -552,6 +555,48 @@ class BasePlugin(ABC):
         """
         return False
 
+    def get_vegas_priority_weight(self) -> Optional[int]:
+        """How many slots per Vegas cycle this plugin should get, or None.
+
+        The Vegas ticker is otherwise a strict round robin: every plugin
+        appears exactly once per cycle. With a dozen plugins enabled that puts
+        minutes between a live score and its next appearance. A weight of N
+        gives the plugin N slots per cycle, spread evenly through it rather
+        than clumped together.
+
+        Return ``None`` (the default) to let the core decide. It gives a
+        plugin ``vegas_scroll.live_weight`` when ``has_live_priority()`` and
+        ``has_live_content()`` are both true, and 1 otherwise -- so live sports
+        already get extra turns without implementing this at all.
+
+        Implement it only when the plugin knows something the core cannot. The
+        motivating case is favorite teams: the core can see *that* a game is
+        live but not *whose*, so a scoreboard that wants its favorite's game
+        shown more often than other live games has to say so::
+
+            def get_vegas_priority_weight(self):
+                if not (self.has_live_priority() and self.has_live_content()):
+                    return None                      # let the core decide
+                cfg = self.global_config.get('display', {}).get('vegas_scroll', {})
+                if self._favorite_is_live():
+                    return cfg.get('favorite_live_weight', 5)
+                return cfg.get('live_weight', 3)
+
+        The weight is per *plugin*, not per game. A scoreboard showing four
+        live games still occupies one slot at a time and rotates its own games
+        within that slot; this controls how often the plugin itself comes
+        round.
+
+        Raising is safe: the core logs it and falls back to its own
+        live-content check, so a broken weight calculation costs the plugin
+        the favorite distinction but not the live boost.
+
+        Returns:
+            Slots per cycle (clamped to 1..10 by the caller), or None to
+            defer to the core's own live-content weighting.
+        """
+        return None
+
     def get_live_modes(self) -> List[str]:
         """
         Get list of display modes that should be used during live priority takeover.
@@ -794,10 +839,12 @@ class BasePlugin(ABC):
                 self.logger.error("'enabled' must be a boolean")
                 return False
 
-        # Check display_duration if present
+        # Check display_duration if present. bool is excluded explicitly:
+        # it's an int subclass, and get_display_duration rejects it too.
         if "display_duration" in self.config:
             duration = self.config["display_duration"]
-            if not isinstance(duration, (int, float)) or duration <= 0:
+            if (not isinstance(duration, (int, float))
+                    or isinstance(duration, bool) or duration <= 0):
                 self.logger.error("'display_duration' must be a positive number")
                 return False
 

src/plugin_system/compatibility.py

@@ -28,6 +28,7 @@ fixes their version string. See `docs/SPORTS_UNIFICATION.md`, phase B4.
 from __future__ import annotations
 
 import re
+from pathlib import Path
 from typing import Any, Dict, Optional, Tuple
 
 # Below this, the core's self-reported version is not evidence of anything.
@@ -35,6 +36,45 @@ from typing import Any, Dict, Optional, Tuple
 TRUSTWORTHY_FLOOR: Tuple[int, int, int] = (2, 0, 0)
 
 
+# ``src/__init__.py`` relative to this file: src/plugin_system/ -> src/
+_VERSION_FILE = Path(__file__).resolve().parent.parent / "__init__.py"
+_VERSION_RE = re.compile(r'^__version__\s*=\s*["\']([^"\']+)["\']', re.M)
+
+
+def current_core_version() -> str:
+    """The core version as it is on disk right now, not as it was at import.
+
+    ``from src import __version__`` binds whatever the process loaded at start.
+    The web UI runs as its own long-lived service (``ledmatrix-web.service``),
+    and updating the core replaces files on disk without restarting it -- the
+    update route says so explicitly and asks the user to restart. Its prompt
+    names the *display* service, so a user who follows it leaves the web
+    process holding the old number.
+
+    The plugin store's gate lives in that web process. Stale by one release is
+    exactly the case that matters: every plugin flooring on the release you
+    just installed gets refused, with a message blaming a core version that is
+    already correct on disk. 3.3.0 is the first release where that hits a whole
+    plugin family at once -- all eight sports scoreboards floor there.
+
+    Reading the file costs one stat and a small read per call, and only on the
+    install/update path. Any failure falls back to the imported value, so this
+    can only ever be as wrong as before, never worse.
+    """
+    try:
+        text = _VERSION_FILE.read_text(encoding="utf-8")
+        match = _VERSION_RE.search(text)
+        if match:
+            return match.group(1)
+    except (OSError, UnicodeDecodeError):
+        pass
+    try:
+        from src import __version__ as imported
+        return imported
+    except Exception:            # noqa: BLE001 - never let this raise
+        return "0.0.0"
+
+
 def parse_semver(value: Any) -> Optional[Tuple[int, int, int]]:
     """Parse ``X.Y.Z`` (extra parts and suffixes ignored) into a comparable
     3-tuple, or ``None`` when unparseable. A leading ``v`` is tolerated."""

src/plugin_system/plugin_health.py

@@ -7,7 +7,7 @@ and circuit breaker state. Provides automatic recovery mechanisms.
 
 import time
 import logging
-from typing import Dict, Optional, Any
+from typing import Dict, Optional, Any, Tuple
 from enum import Enum
 
 
@@ -64,10 +64,48 @@ class PluginHealthTracker:
             cache_key, max_age=None, memory_ttl=0 if force_reload else None
         )
 
-        if cached:
+        if isinstance(cached, dict) and cached:
-            return cached
+            # Complete it rather than trusting it: a persisted record can be
+            # missing fields the callers index directly (a partial write, a
+            # restored backup, an older schema), and returning it verbatim makes
+            # record_success / record_failure raise KeyError, which takes the
+            # display down in a restart loop that survives reboots because the
+            # bad entry is on disk.
+            state, repaired = self._repair_health_state(cached)
+            if repaired:
+                self.logger.warning(
+                    f"Repaired health state for {plugin_id}: "
+                    f"{sorted(repaired)} missing or invalid, using defaults for those."
+                )
+            return state
 
-        # Default state
+        # Not a dict at all: written by something other than
+        # _save_health_state (a key collision, a corrupted entry). Nothing to
+        # salvage.
+        if cached is not None and not isinstance(cached, dict):
+            self.logger.warning(
+                f"Discarding malformed health state for {plugin_id}: expected "
+                f"dict, got {type(cached).__name__}. Falling back to defaults."
+            )
+
+        return self._default_health_state()
+    
+    def _save_health_state(self, plugin_id: str, state: Dict[str, Any]) -> None:
+        """Save health state to cache."""
+        cache_key = self._get_health_key(plugin_id)
+        self.cache_manager.set(cache_key, state)  # Persist indefinitely
+        self._health_state[plugin_id] = state
+    
+    # The fields callers index directly (state['circuit_state'] and friends).
+    # A cached dict missing any of them raises KeyError deep in record_success /
+    # record_failure, so the value is completed before it is handed out.
+    _COUNTER_FIELDS = ('consecutive_failures', 'total_failures', 'total_successes')
+    _TIMESTAMP_FIELDS = ('last_success_time', 'last_failure_time',
+                         'circuit_opened_time', 'half_open_start_time')
+
+    @staticmethod
+    def _default_health_state() -> Dict[str, Any]:
+        """A fresh state with every field the callers expect."""
         return {
             'consecutive_failures': 0,
             'total_failures': 0,
@@ -77,14 +115,55 @@ class PluginHealthTracker:
             'circuit_state': CircuitState.CLOSED.value,
             'circuit_opened_time': None,
             'half_open_start_time': None,
-            'last_error': None
+            'last_error': None,
         }
 
-    def _save_health_state(self, plugin_id: str, state: Dict[str, Any]) -> None:
+    @classmethod
-        """Save health state to cache."""
+    def _repair_health_state(cls, cached: Dict[str, Any]) -> Tuple[Dict[str, Any], list]:
-        cache_key = self._get_health_key(plugin_id)
+        """Return `cached` completed against the defaults, plus what was repaired.
-        self.cache_manager.set(cache_key, state)  # Persist indefinitely
+
-        self._health_state[plugin_id] = state
+        Per-field rather than all-or-nothing: a record that has real failure
+        counts but is missing `last_error` should keep the counts, not be reset
+        to healthy. Only values that are absent or the wrong type fall back to
+        the default, so a partial or older-schema record survives with whatever
+        it does carry, while every field the callers index is guaranteed present
+        and of a usable type.
+        """
+        state = cls._default_health_state()
+        repaired = []
+        for field, default in state.items():
+            if field not in cached:
+                repaired.append(field)
+                continue
+            value = cached[field]
+            if field in cls._COUNTER_FIELDS:
+                ok = isinstance(value, int) and not isinstance(value, bool) and value >= 0
+            elif field in cls._TIMESTAMP_FIELDS:
+                # bool is a subclass of int, so True would pass as a timestamp
+                # and then compare as 1.0 -- expiring a cooldown the instant it
+                # opens, or (False) making the elapsed check never fire.
+                ok = value is None or (
+                    isinstance(value, (int, float)) and not isinstance(value, bool)
+                )
+            elif field == 'circuit_state':
+                # Membership first requires the value to be hashable: a list or
+                # dict here would raise TypeError out of the repair itself,
+                # which is the crash this whole path exists to prevent.
+                ok = isinstance(value, str) and value in {
+                    member.value for member in CircuitState
+                }
+            else:  # last_error
+                ok = value is None or isinstance(value, str)
+            if ok:
+                state[field] = value
+            else:
+                repaired.append(field)
+        # Anything the schema has since grown (degraded, degraded_reason) is
+        # read with .get() by its callers, so carry it through untouched.
+        for field, value in cached.items():
+            if field not in state:
+                state[field] = value
+        return state, repaired
 
     def get_health_state(self, plugin_id: str, force_reload: bool = False) -> Dict[str, Any]:
         """Get current health state for a plugin.
@@ -99,10 +178,20 @@ class PluginHealthTracker:
             )
         return self._health_state[plugin_id]
     
+    # Fields the circuit breaker is rebuilt from after a restart. Everything
+    # else in a health record is reporting, read only for display.
+    _DURABLE_FIELDS = ('consecutive_failures', 'circuit_state',
+                       'circuit_opened_time', 'half_open_start_time')
+
+    def _durable(self, state: Dict[str, Any]) -> tuple:
+        """The part of a health record whose loss would change behaviour."""
+        return tuple(state.get(field) for field in self._DURABLE_FIELDS)
+
     def record_success(self, plugin_id: str) -> None:
         """Record a successful plugin execution."""
         state = self.get_health_state(plugin_id)
         current_time = time.time()
+        durable_before = self._durable(state)
 
         # Reset consecutive failures
         state['consecutive_failures'] = 0
@@ -120,6 +209,17 @@ class PluginHealthTracker:
             state['circuit_state'] = CircuitState.CLOSED.value
             state['circuit_opened_time'] = None
 
+        # A healthy plugin reports success every cycle, and in that steady state
+        # the only fields changed above are a counter and a timestamp that
+        # nothing reads back after a restart. Persisting them anyway rewrites a
+        # small file per plugin per cycle: on a rig running 24 plugins, a
+        # five-minute sample measured 22 rewrites, about 4.4 a minute or 6,300 a
+        # day. Those land on an SD card, where the cost is an erase-block cycle
+        # rather than the 400 bytes involved, and where wear is what eventually
+        # kills the card.
+        # In-memory state is still updated every time, so the health API and web
+        # UI show exactly what they did before; only the write is skipped.
+        if self._durable(state) != durable_before:
             self._save_health_state(plugin_id, state)
 
     def record_failure(self, plugin_id: str, error: Optional[Exception] = None) -> None:

src/plugin_system/plugin_loader.py

@@ -14,7 +14,7 @@ import sys
 import subprocess
 import threading
 from pathlib import Path
-from typing import Dict, Any, Optional, Tuple, Type
+from typing import Dict, Any, List, Optional, Tuple, Type
 import logging
 
 from packaging.requirements import InvalidRequirement, Requirement
@@ -45,6 +45,76 @@ def requirements_has_real_deps(requirements_file: str) -> bool:
     return False
 
 
+def _extra_dependencies(dist_name: str, extras) -> Optional[List[Requirement]]:
+    """Dependencies a distribution declares *only* behind the given extras.
+
+    Returns None when the installed metadata cannot be read or parsed, so the
+    caller can fall back to running pip rather than assuming anything.
+    """
+    try:
+        meta = importlib.metadata.metadata(dist_name)
+    except importlib.metadata.PackageNotFoundError:
+        return None
+
+    gated: List[Requirement] = []
+    for raw in meta.get_all('Requires-Dist') or []:
+        try:
+            dep = Requirement(raw)
+        except InvalidRequirement:
+            return None
+        if dep.marker is None:
+            continue
+        # Keep only what the distribution gates behind an extra we asked for:
+        # satisfied when `extra` is that name, but not when no extra is
+        # requested. A marker that holds either way (python_version, sys_platform)
+        # belongs to the base install and is already covered by the version check.
+        if dep.marker.evaluate({'extra': ''}):
+            continue
+        if any(dep.marker.evaluate({'extra': extra}) for extra in extras):
+            gated.append(dep)
+    return gated
+
+
+def _extras_are_satisfied(req: Requirement, _visited: Optional[set] = None) -> bool:
+    """Check the dependencies pulled in by req's extras are installed.
+
+    Follows extras through nested extras. A gated dependency can itself request
+    one (`requests[socks]`), and checking only that `requests` is installed at
+    an acceptable version says nothing about whether the socks extra's own
+    dependency is there -- so the caller would skip pip and the plugin would
+    fail at import instead. Plain dependencies are still checked one level
+    deep, which is all that is needed to tell "the extra was installed" from
+    "the extra was never installed".
+
+    `_visited` carries the (distribution, extras) pairs already seen, so a
+    dependency cycle between extras terminates instead of recursing forever.
+    Anything unreadable returns False, so the caller still falls through to pip.
+    """
+    if _visited is None:
+        _visited = set()
+    marker = (req.name.lower(), frozenset(e.lower() for e in req.extras))
+    if marker in _visited:
+        # Already accounted for higher up the chain; treating a cycle as
+        # satisfied here is safe because the outer frame still has to pass.
+        return True
+    _visited.add(marker)
+
+    gated = _extra_dependencies(req.name, req.extras)
+    if gated is None:
+        return False
+
+    for dep in gated:
+        try:
+            dep_version = importlib.metadata.version(dep.name)
+        except importlib.metadata.PackageNotFoundError:
+            return False
+        if dep.specifier and not dep.specifier.contains(dep_version, prereleases=True):
+            return False
+        if dep.extras and not _extras_are_satisfied(dep, _visited):
+            return False
+    return True
+
+
 def requirements_are_satisfied(requirements_file: str) -> bool:
     """
     Check whether every real requirement line in requirements.txt is already
@@ -76,9 +146,6 @@ def requirements_are_satisfied(requirements_file: str) -> bool:
         except InvalidRequirement:
             return False
 
-        if req.extras:
-            return False  # verifying extras' sub-dependencies isn't worth it here
-
         if req.marker is not None and not req.marker.evaluate():
             continue  # not applicable on this platform/interpreter
 
@@ -90,6 +157,9 @@ def requirements_are_satisfied(requirements_file: str) -> bool:
         if req.specifier and not req.specifier.contains(installed_version, prereleases=True):
             return False
 
+        if req.extras and not _extras_are_satisfied(req):
+            return False
+
     return True
 
 
@@ -702,8 +772,8 @@ class PluginLoader:
         newer than the running core. Advisory only — never raises — so a
         plugin that guards optional features with try/except keeps working.
         """
-        from src import __version__ as core_version
         from src.plugin_system import compatibility
+        core_version = compatibility.current_core_version()
 
         compatible, _reason = compatibility.check(manifest, core_version)
         if compatible:

src/plugin_system/plugin_manager.py

@@ -71,11 +71,15 @@ class PluginManager:
         self.plugin_loader = PluginLoader(logger=self.logger)
         self.plugin_executor = PluginExecutor(default_timeout=30.0, logger=self.logger)
         self.state_manager = PluginStateManager(logger=self.logger)
-        self.schema_manager = SchemaManager(plugins_dir=self.plugins_dir, logger=self.logger)
+        self.schema_manager = SchemaManager(plugins_dir=self.plugins_dir, logger=self.logger,
+                                           config_manager=self.config_manager)
         
         # Lock protecting plugin_manifests and plugin_directories from
         # concurrent mutation (background reconciliation) and reads (requests).
         self._discovery_lock = threading.RLock()
+        #: Directories already reported as unloadable, so the warning is
+        #: emitted once rather than on every discovery scan.
+        self._skip_reported: set = set()
 
         # Lock protecting plugin_last_update from concurrent mutation/iteration.
         # It's written from run_scheduled_updates()/update_all_plugins() (main
@@ -125,6 +129,14 @@ class PluginManager:
         self._plugin_locks: Dict[str, threading.Lock] = {}
         self._plugin_locks_guard = threading.Lock()
         self._update_worker: Optional[threading.Thread] = None
+        # Plugin ids whose update() has finished since the last time anyone
+        # asked. Updates are dispatched to a worker thread, so a caller that
+        # wants to know "whose data just changed" cannot learn it by diffing
+        # plugin_last_update around run_scheduled_updates() -- that call only
+        # enqueues, and the timestamp is stamped later, on the worker. See
+        # run_scheduled_updates_with_changes().
+        self._completed_updates: set = set()
+        self._completed_updates_lock = threading.Lock()
         self._synchronous_updates = False
         if self.config_manager is not None:
             try:
@@ -187,18 +199,59 @@ class PluginManager:
                     continue
 
                 manifest_path = item / "manifest.json"
-                if manifest_path.exists():
+                if not manifest_path.exists():
+                    # Once per directory per process. Discovery runs on every
+                    # web UI page load and every config reconcile, so warning
+                    # unconditionally would put a line in the journal each
+                    # time someone opened a page -- the same log-volume
+                    # problem this is meant to help diagnose.
+                    # A directory here that carries no manifest is not a
+                    # plugin. Said once, because the alternative is a plugin
+                    # that is enabled in config, enabled in plugin state,
+                    # present on disk, and simply absent from the running
+                    # process with nothing anywhere to say why. Working that
+                    # out afterwards means reading cache-file mtimes.
+                    if item.name not in self._skip_reported:
+                        self._skip_reported.add(item.name)
+                        self.logger.warning(
+                            "Skipping %s: no manifest.json, so it cannot be "
+                            "loaded as a plugin", item.name)
+                    continue
                 try:
                     with open(manifest_path, 'r', encoding='utf-8') as f:
                         manifest = json.load(f)
-                            plugin_id = manifest.get('id')
-                            if plugin_id:
-                                plugin_ids.append(plugin_id)
-                                new_manifests[plugin_id] = manifest
-                                new_directories[plugin_id] = item
                 except (json.JSONDecodeError, PermissionError, OSError) as e:
                     self.logger.warning("Error reading manifest from %s: %s", manifest_path, e, exc_info=True)
                     continue
+
+                # json.load accepts any JSON value, so a manifest holding
+                # null, [] or "text" parses and then raises AttributeError on
+                # .get(). Nothing here catches that -- the outer handler takes
+                # OSError/PermissionError only -- so a single malformed
+                # manifest aborted the whole scan and every other plugin on
+                # disk, however healthy, silently failed to register.
+                if not isinstance(manifest, dict):
+                    if item.name not in self._skip_reported:
+                        self._skip_reported.add(item.name)
+                        self.logger.warning(
+                            "Skipping %s: its manifest.json is %s, not a JSON "
+                            "object", item.name, type(manifest).__name__)
+                    continue
+
+                plugin_id = manifest.get('id')
+                if not plugin_id:
+                    # Parsed but unusable. This was the quietest path of all:
+                    # the manifest is read successfully and then dropped.
+                    if item.name not in self._skip_reported:
+                        self._skip_reported.add(item.name)
+                        self.logger.warning(
+                            "Skipping %s: its manifest.json has no \"id\", so "
+                            "there is nothing to register it under", item.name)
+                    continue
+
+                plugin_ids.append(plugin_id)
+                new_manifests[plugin_id] = manifest
+                new_directories[plugin_id] = item
         except (OSError, PermissionError) as e:
             self.logger.error("Error scanning directory %s: %s", directory, e, exc_info=True)
 
@@ -578,6 +631,17 @@ class PluginManager:
         
         return self.load_plugin(plugin_id)
     
+    def discovered_plugin_ids(self) -> set:
+        """Snapshot of the discovered plugin ids, taken under the discovery lock.
+
+        Callers on other threads (the config watcher) must not iterate
+        ``plugin_manifests`` directly: discovery rebuilds it entry by entry, so
+        an unsynchronised reader can see a half-populated mapping or raise
+        "dictionary changed size during iteration".
+        """
+        with self._discovery_lock:
+            return set(self.plugin_manifests)
+
     def get_plugin(self, plugin_id: str) -> Optional[Any]:
         """
         Get a loaded plugin instance by ID.
@@ -1025,6 +1089,7 @@ class PluginManager:
                 if success:
                     with self._plugin_last_update_lock:
                         self.plugin_last_update[plugin_id] = scheduled_time
+                    self._note_update_completed(plugin_id)
                     self.state_manager.record_update(plugin_id)
                     self.state_manager.set_state(plugin_id, PluginState.ENABLED)
                     if self.health_tracker:
@@ -1089,28 +1154,41 @@ class PluginManager:
 
     def run_scheduled_updates_with_changes(self, current_time: Optional[float] = None) -> List[str]:
         """
-        Like run_scheduled_updates(), but also returns the plugin_ids whose
+        Like run_scheduled_updates(), but also reports which plugins have
-        plugin_last_update timestamp actually advanced during this call.
+        fresh data -- the ids whose update() has finished since the last
+        call, not necessarily the ones enqueued by this one.
 
-        The before/after snapshots and the update pass itself are each
+        That distinction is the whole point. This used to snapshot
-        individually lock-protected against concurrent plugin_last_update
+        plugin_last_update, call run_scheduled_updates(), and diff. But
-        mutation (Vegas mode calls this from its own background
+        run_scheduled_updates() only *enqueues*: the work runs on the
-        update-tick thread, racing the main render loop's plugin updates),
+        update worker and the timestamp is stamped there, after this method
-        so callers get an atomic "who got fresh data" answer without
+        has already returned. The two snapshots were therefore always
-        reaching into plugin_last_update themselves. The lock is not held
+        identical and the result was always empty, so Vegas never learned
-        across the update pass so slow/blocking plugin update() calls don't
+        that any plugin's data had changed and kept scrolling whatever a
-        serialize against other plugin_last_update readers.
+        segment was first built from -- last night's live game still drawn
+        as live the next morning. The only path that ever worked was the
+        synchronous kill-switch, where update() runs inline.
+
+        Reporting completions instead of enqueues costs a poll's worth of
+        latency (the Vegas tick runs every ~4s) and is correct regardless of
+        which side of the queue the work lands on.
         """
-        with self._plugin_last_update_lock:
-            old_times = dict(self.plugin_last_update)
-
         self.run_scheduled_updates(current_time)
+        return self.drain_completed_updates()
 
-        with self._plugin_last_update_lock:
+    def _note_update_completed(self, plugin_id: str) -> None:
-            return [
+        """Record that a plugin's update() finished, for the next poll."""
-                plugin_id for plugin_id, new_time in self.plugin_last_update.items()
+        with self._completed_updates_lock:
-                if new_time > old_times.get(plugin_id, 0.0)
+            self._completed_updates.add(plugin_id)
-            ]
+
+    def drain_completed_updates(self) -> List[str]:
+        """Return and clear the plugin ids whose update() has since finished."""
+        with self._completed_updates_lock:
+            if not self._completed_updates:
+                return []
+            done = sorted(self._completed_updates)
+            self._completed_updates.clear()
+            return done
 
     def update_all_plugins(self) -> None:
         """
@@ -1135,6 +1213,7 @@ class PluginManager:
                 if success:
                     with self._plugin_last_update_lock:
                         self.plugin_last_update[plugin_id] = time.time()
+                    self._note_update_completed(plugin_id)
                     self.state_manager.record_update(plugin_id)
                     self.state_manager.set_state(plugin_id, PluginState.ENABLED)
                 else:

src/plugin_system/plugin_state.py

@@ -6,14 +6,40 @@ with state transitions and queries.
 """
 
 import threading
+import time
+from collections import deque
 from enum import Enum
-from typing import Optional, Dict, Any
+from typing import Optional, Dict, Any, Deque, List, Tuple
 from datetime import datetime
 import logging
 
 from src.logging_config import get_logger
 
 
+# The history is diagnostic only -- nothing reads the entries themselves, just
+# their count -- but it is appended to on the hot scheduling path: every update
+# cycle records RUNNING on reserve and ENABLED on finish. Unbounded, that is
+# 2,880 entries per plugin per day at the default 60s interval, which on a 1 GB
+# Pi exhausts memory in weeks.
+#
+# Two limits, because a single entry count answers the wrong question. What a
+# reader wants is "the last couple of hours", and how many transitions that is
+# depends entirely on the plugin's update interval -- which on a real board
+# spans 2s to 3600s. A flat 200 entries is 4.2 days for the slowest plugin and
+# 3.3 minutes for the fastest, so the plugin churning hardest, the one worth
+# looking at, keeps the least history.
+#
+# So: trim by AGE first, which makes the retained window comparable across
+# plugins whatever their cadence...
+STATE_HISTORY_MAX_AGE_SECONDS = 2 * 60 * 60
+
+# ...and cap by COUNT second, purely as a memory ceiling for the fast pollers
+# whose age window would otherwise run to thousands of entries. At ~230 bytes
+# an entry this is ~0.5 MB per plugin worst case, and only plugins updating
+# faster than roughly every 4s can reach it.
+MAX_STATE_HISTORY_PER_PLUGIN = 2000
+
+
 class PluginState(Enum):
     """Plugin state enumeration."""
     UNLOADED = "unloaded"  # Plugin not loaded
@@ -37,11 +63,43 @@ class PluginStateManager:
         self.logger = logger or get_logger(__name__)
         self._lock = threading.RLock()
         self._states: Dict[str, PluginState] = {}
-        self._state_history: Dict[str, list] = {}
+        # (monotonic timestamp, transition). The clock is monotonic so a DST
+        # shift or an NTP step cannot make entries look old and flush the
+        # history; the human-readable timestamp lives inside the transition.
+        self._state_history: Dict[str, Deque[Tuple[float, Dict[str, Any]]]] = {}
+        # Lifetime transition totals, kept separately so the count reported by
+        # get_state_info() stays truthful once the history above starts rolling.
+        self._state_transition_counts: Dict[str, int] = {}
         self._error_info: Dict[str, Dict[str, Any]] = {}
         self._last_update: Dict[str, datetime] = {}
         self._last_display: Dict[str, datetime] = {}
     
+    def _record_transition(
+        self,
+        plugin_id: str,
+        transition: Dict[str, Any]
+    ) -> None:
+        """Append a transition to the plugin's bounded history.
+
+        Callers must already hold ``_lock``. The deque discards its oldest
+        entry once it is full, so the history cannot grow without bound; the
+        lifetime total is tracked separately for get_state_info().
+        """
+        history = self._state_history.get(plugin_id)
+        if history is None:
+            history = deque(maxlen=MAX_STATE_HISTORY_PER_PLUGIN)
+            self._state_history[plugin_id] = history
+        now = time.monotonic()
+        history.append((now, transition))
+        # Age out first; the deque's maxlen is the backstop for plugins that
+        # produce more than the ceiling within the window.
+        cutoff = now - STATE_HISTORY_MAX_AGE_SECONDS
+        while history and history[0][0] < cutoff:
+            history.popleft()
+        self._state_transition_counts[plugin_id] = (
+            self._state_transition_counts.get(plugin_id, 0) + 1
+        )
+
     def set_state(
         self,
         plugin_id: str,
@@ -60,16 +118,13 @@ class PluginStateManager:
             old_state = self._states.get(plugin_id, PluginState.UNLOADED)
             self._states[plugin_id] = state
 
-            if plugin_id not in self._state_history:
-                self._state_history[plugin_id] = []
-
             transition = {
                 'timestamp': datetime.now(),
                 'from': old_state.value,
                 'to': state.value,
                 'error': str(error) if error else None
             }
-            self._state_history[plugin_id].append(transition)
+            self._record_transition(plugin_id, transition)
 
             # Store error info if transitioning to ERROR state
             if state == PluginState.ERROR and error:
@@ -126,17 +181,29 @@ class PluginStateManager:
         state = self.get_state(plugin_id)
         return state == PluginState.ENABLED
     
-    def get_state_history(self, plugin_id: str) -> list:
+    def get_state_history(self, plugin_id: str) -> List[Dict[str, Any]]:
         """
         Get state transition history for a plugin.
 
+        Retention is by age first -- transitions older than
+        STATE_HISTORY_MAX_AGE_SECONDS are dropped -- and by count second, at
+        MAX_STATE_HISTORY_PER_PLUGIN, which only binds for plugins updating
+        fast enough to exceed it inside that window.
+
         Args:
             plugin_id: Plugin identifier
 
         Returns:
-            List of state transitions
+            List of recent state transitions, oldest first. Both the list and
+            the transition dicts are copies, so callers cannot mutate the
+            manager's own history. The values inside a transition are all
+            immutable, so a shallow copy per entry is enough.
         """
-        return self._state_history.get(plugin_id, [])
+        with self._lock:
+            return [
+                dict(transition)
+                for _stamp, transition in self._state_history.get(plugin_id, ())
+            ]
     
     def set_error_info(self, plugin_id: str, error_info: Dict[str, Any]) -> None:
         """
@@ -179,9 +246,7 @@ class PluginStateManager:
             old_state = self._states.get(plugin_id, PluginState.UNLOADED)
             self._states[plugin_id] = state
 
-            if plugin_id not in self._state_history:
+            self._record_transition(plugin_id, {
-                self._state_history[plugin_id] = []
-            self._state_history[plugin_id].append({
                 'timestamp': datetime.now(),
                 'from': old_state.value,
                 'to': state.value,
@@ -241,6 +306,13 @@ class PluginStateManager:
         Returns:
             Dictionary with state information
         """
+        # One snapshot, one critical section. Each field was read under its own
+        # lock, so an unload running concurrently could be observed half-done:
+        # 'state' read before clear_state() removed it and
+        # 'state_history_count' read after, giving a caller a plugin that is
+        # ENABLED with zero transitions. _lock is an RLock, so the helpers
+        # below can still take it.
+        with self._lock:
             state = self.get_state(plugin_id)
             info = {
                 'state': state.value,
@@ -252,14 +324,21 @@ class PluginStateManager:
                 'last_update': self.get_last_update(plugin_id),
                 'last_display': self.get_last_display(plugin_id),
                 'error_info': self.get_error_info(plugin_id),
-            'state_history_count': len(self.get_state_history(plugin_id))
+                'state_history_count': self._state_transition_counts.get(plugin_id, 0)
             }
         return info
     
     def clear_state(self, plugin_id: str) -> None:
-        """Clear all state information for a plugin."""
+        """Clear all state information for a plugin.
+
+        Held under ``_lock`` so the five dicts are dropped as one unit: every
+        other mutator takes the lock, and without it a concurrent set_state()
+        could interleave and leave a plugin with history but no state.
+        """
+        with self._lock:
             self._states.pop(plugin_id, None)
             self._state_history.pop(plugin_id, None)
+            self._state_transition_counts.pop(plugin_id, None)
             self._error_info.pop(plugin_id, None)
             self._last_update.pop(plugin_id, None)
             self._last_display.pop(plugin_id, None)

src/plugin_system/resource_monitor.py

@@ -9,7 +9,7 @@ import time
 import logging
 import threading
 from typing import Dict, Optional, Any, Callable
-from dataclasses import dataclass, field
+from dataclasses import dataclass, field, fields
 
 try:
     import psutil
@@ -49,6 +49,20 @@ class ResourceMetrics:
             self.total_execution_time = self.total_execution_time / self.call_count
 
 
+#: How often a plugin's metrics are written to the cache, in seconds.
+#:
+#: Persisting on every call meant a small file rewritten roughly nine times a
+#: minute per plugin. On a rig with fourteen active plugins that was ~126
+#: writes a minute for metrics alone, and since each ~350-byte file costs a
+#: 4KB block plus an ext4 journal entry, it dominated the device's write
+#: volume -- on an SD card, which wears out.
+#:
+#: The in-memory copy stays authoritative and exact; only the cross-process
+#: snapshot the web UI reads is delayed, and telemetry up to half a minute old
+#: is still a fair description of a long-running plugin.
+_METRICS_PERSIST_INTERVAL = 30.0
+
+
 class PluginResourceMonitor:
     """
     Monitors resource usage for plugins.
@@ -75,6 +89,10 @@ class PluginResourceMonitor:
         # Resource metrics per plugin
         self._metrics: Dict[str, ResourceMetrics] = {}
         self._limits: Dict[str, ResourceLimits] = {}
+        # When each plugin's metrics last reached the cache. Metrics change on
+        # every call, so they cannot be de-duplicated the way health state can;
+        # they are rate-limited instead. See _METRICS_PERSIST_INTERVAL.
+        self._metrics_persisted_at: Dict[str, float] = {}
 
         # Thread-local storage for execution tracking
         self._local = threading.local()
@@ -102,6 +120,66 @@ class PluginResourceMonitor:
                 "psutil not available - resource monitoring will be limited to execution time only"
             )
     
+    def _metrics_from_cache(self, plugin_id: str, cached: Any) -> "ResourceMetrics":
+        """Build metrics from a cached record, ignoring anything unrecognised.
+
+        ResourceMetrics(**cached) raises TypeError on a single unexpected key,
+        and that exception escapes into plugin_manager, which reports it as
+        "plugin <id> operation failed". Every plugin fails, and the plugin
+        system never finishes initialising.
+
+        Seen on a live rig: every plugin failing with
+
+            ResourceMetrics.__init__() got an unexpected keyword argument
+            'consecutive_failures'
+
+        which is a plugin_health field, not a metrics one. How a health-shaped
+        record came to sit under a plugin_metrics key on that machine is not
+        established -- a restored backup that mixed two machines' caches is the
+        likeliest explanation -- but the loader should not be brittle enough for
+        it to matter. plugin_health already repairs its records field by field
+        rather than trusting whatever is on disk; this does the same.
+
+        Unknown keys are dropped and named once, so a genuine schema change is
+        visible in the log instead of silently discarded.
+        """
+        if not isinstance(cached, dict):
+            self.logger.warning(
+                "Ignoring cached metrics for %s: expected a mapping, got %s",
+                plugin_id, type(cached).__name__)
+            return ResourceMetrics()
+
+        known = {f.name for f in fields(ResourceMetrics)}
+        unknown = sorted(set(cached) - known)
+        if unknown:
+            self.logger.warning(
+                "Dropping unrecognised field(s) from cached metrics for %s: %s",
+                plugin_id, ", ".join(unknown))
+        # A dataclass does not enforce its annotations, so
+        # ResourceMetrics(call_count="not a number") builds happily and only
+        # blows up later, deep inside monitor_call ("can only concatenate str
+        # (not \"int\") to str"). Coerce here, where there is still a cache
+        # key to name in the warning.
+        declared = {f.name: f.type for f in fields(ResourceMetrics)}
+        usable = {}
+        for key, value in cached.items():
+            if key not in known:
+                continue
+            try:
+                usable[key] = int(value) if declared[key] in ('int', int) else float(value)
+            except (TypeError, ValueError):
+                self.logger.warning(
+                    "Cached metrics for %s have a bad %s (%r); starting fresh",
+                    plugin_id, key, value)
+                return ResourceMetrics()
+        try:
+            return ResourceMetrics(**usable)
+        except (TypeError, ValueError) as e:
+            self.logger.warning(
+                "Cached metrics for %s unusable (%s); starting fresh",
+                plugin_id, e)
+            return ResourceMetrics()
+
     def _get_metrics_key(self, plugin_id: str) -> str:
         """Get cache key for plugin metrics."""
         return f"plugin_metrics:{plugin_id}"
@@ -126,7 +204,7 @@ class PluginResourceMonitor:
                     cache_key, max_age=None, memory_ttl=0 if force_reload else None
                 )
                 if cached:
-                    metrics = ResourceMetrics(**cached)
+                    metrics = self._metrics_from_cache(plugin_id, cached)
                 else:
                     metrics = ResourceMetrics()
                 self._metrics[plugin_id] = metrics
@@ -232,18 +310,8 @@ class PluginResourceMonitor:
                     # CPU is harder to measure per-call, so we track it separately
                     metrics.cpu_percent = self._get_process_cpu_percent()
                 
-                # Persist metrics
+                # Persist metrics, at most once per interval per plugin.
-                cache_key = self._get_metrics_key(plugin_id)
+                self._persist_metrics(plugin_id, metrics)
-                self.cache_manager.set(cache_key, {
-                    'memory_mb': metrics.memory_mb,
-                    'cpu_percent': metrics.cpu_percent,
-                    'execution_time': metrics.execution_time,
-                    'call_count': metrics.call_count,
-                    'total_execution_time': metrics.total_execution_time,
-                    'max_execution_time': metrics.max_execution_time,
-                    'min_execution_time': metrics.min_execution_time if metrics.min_execution_time != float('inf') else 0.0,
-                    'last_update_time': metrics.last_update_time
-                })
             
             # Check limits
             if limits:
@@ -363,6 +431,44 @@ class PluginResourceMonitor:
             summaries[plugin_id] = self.get_metrics_summary(plugin_id)
         return summaries
     
+    def _persist_metrics(self, plugin_id: str, metrics: ResourceMetrics,
+                         force: bool = False) -> None:
+        """Write a plugin's metrics to the cache, at most once per interval.
+
+        Caller must hold ``self._lock``.
+        """
+        # Monotonic, not wall clock: these devices have no RTC, so the clock
+        # jumps by however far off boot-time was the moment NTP first syncs.
+        # A forward jump would allow an early write, a backward one would
+        # stall the snapshot well past the interval.
+        #
+        # The sentinel for "never written" is None, not 0.0. monotonic() is
+        # time since boot on Linux, and systemd starts this service *at* boot,
+        # so `now - 0.0 < 30` was true for the first half-minute of every
+        # single run -- the throttle swallowed the very first snapshot, which
+        # is the one that matters most after a restart.
+        now = time.monotonic()
+        last_written = self._metrics_persisted_at.get(plugin_id)
+        if (not force and last_written is not None
+                and now - last_written < _METRICS_PERSIST_INTERVAL):
+            return
+        cache_key = self._get_metrics_key(plugin_id)
+        self.cache_manager.set(cache_key, {
+            'memory_mb': metrics.memory_mb,
+            'cpu_percent': metrics.cpu_percent,
+            'execution_time': metrics.execution_time,
+            'call_count': metrics.call_count,
+            'total_execution_time': metrics.total_execution_time,
+            'max_execution_time': metrics.max_execution_time,
+            'min_execution_time': (metrics.min_execution_time
+                                   if metrics.min_execution_time != float('inf')
+                                   else 0.0),
+            'last_update_time': metrics.last_update_time,
+        })
+        # Only after the write lands. Marking it first would mean a failed
+        # set() bought the next interval's silence without leaving a snapshot.
+        self._metrics_persisted_at[plugin_id] = now
+
     def reset_metrics(self, plugin_id: str) -> None:
         """Reset metrics for a plugin."""
         with self._lock:
@@ -370,4 +476,7 @@ class PluginResourceMonitor:
                 self._metrics[plugin_id] = ResourceMetrics()
                 cache_key = self._get_metrics_key(plugin_id)
                 self.cache_manager.delete(cache_key)
+                # Let the next call persist immediately rather than leaving the
+                # deleted key absent for the rest of the interval.
+                self._metrics_persisted_at.pop(plugin_id, None)
 

src/plugin_system/saved_repositories.py

@@ -6,6 +6,7 @@ Manages saved GitHub repository URLs for easy plugin discovery and installation.
 
 import json
 import logging
+import os
 from pathlib import Path
 from typing import List, Dict, Optional
 
@@ -43,20 +44,45 @@ class SavedRepositoriesManager:
             return []
     
     def _save_repositories(self) -> bool:
-        """Save repositories to file."""
+        """Save repositories to file atomically.
+
+        Writes to a temp file in the same directory and os.replace()s it
+        over the target, so a failed write can never truncate or
+        half-overwrite an existing saved_repositories.json.
+        """
+        tmp_path = self.config_path.with_suffix(self.config_path.suffix + '.tmp')
         try:
             # Ensure directory exists
             self.config_path.parent.mkdir(parents=True, exist_ok=True)
 
-            with open(self.config_path, 'w') as f:
+            with open(tmp_path, 'w') as f:
                 json.dump(self.repositories, f, indent=2)
+                f.flush()
+                os.fsync(f.fileno())
+            os.replace(tmp_path, self.config_path)
 
             self.logger.info(f"Saved {len(self.repositories)} repositories to {self.config_path}")
             return True
         except Exception as e:
             self.logger.error(f"Error saving repositories: {e}")
+            try:
+                tmp_path.unlink(missing_ok=True)
+            except OSError:
+                pass
             return False
     
+    @staticmethod
+    def _clean_url(repo_url: str) -> str:
+        """Normalize a repo URL: strip whitespace, trailing slashes, and a
+        trailing ``.git`` suffix ONLY. (The old ``.replace('.git', '')``
+        was an unanchored substring replace that mangled URLs merely
+        containing ``.git``, e.g. ``https://github.com/user/my.github.io``.)
+        """
+        repo_url = repo_url.strip().rstrip('/')
+        if repo_url.endswith('.git'):
+            repo_url = repo_url[:-4]
+        return repo_url
+
     def get_all(self) -> List[Dict[str, str]]:
         """Get all saved repositories."""
         return self.repositories.copy()
@@ -72,8 +98,7 @@ class SavedRepositoriesManager:
         Returns:
             True if added successfully
         """
-        # Clean URL
+        repo_url = self._clean_url(repo_url)
-        repo_url = repo_url.strip().rstrip('/').replace('.git', '')
 
         # Check if already exists
         for repo in self.repositories:
@@ -96,7 +121,12 @@ class SavedRepositoriesManager:
             'type': 'registry' if 'plugins.json' in repo_url or 'ledmatrix-plugins' in repo_url.lower() else 'single'
         })
 
-        return self._save_repositories()
+        if not self._save_repositories():
+            # Keep memory consistent with disk: a failed save must not leave
+            # a phantom entry that only this process can see.
+            self.repositories.pop()
+            return False
+        return True
     
     def remove(self, repo_url: str) -> bool:
         """
@@ -108,21 +138,25 @@ class SavedRepositoriesManager:
         Returns:
             True if removed successfully
         """
-        # Clean URL
+        repo_url = self._clean_url(repo_url)
-        repo_url = repo_url.strip().rstrip('/').replace('.git', '')
 
-        original_count = len(self.repositories)
+        previous = self.repositories
-        self.repositories = [r for r in self.repositories if r.get('url') != repo_url]
+        remaining = [r for r in previous if r.get('url') != repo_url]
 
-        if len(self.repositories) < original_count:
+        if len(remaining) < len(previous):
-            return self._save_repositories()
+            self.repositories = remaining
+            if not self._save_repositories():
+                # Failed save: restore so memory matches disk.
+                self.repositories = previous
+                return False
+            return True
         else:
             self.logger.warning(f"Repository not found: {repo_url}")
             return False
     
     def has(self, repo_url: str) -> bool:
         """Check if a repository is already saved."""
-        repo_url = repo_url.strip().rstrip('/').replace('.git', '')
+        repo_url = self._clean_url(repo_url)
         return any(r.get('url') == repo_url for r in self.repositories)
     
     def get_registry_repositories(self) -> List[Dict[str, str]]:

src/plugin_system/schema_manager.py

@@ -26,7 +26,25 @@ class SchemaManager:
     - Cache invalidation on plugin changes
     """
     
-    def __init__(self, plugins_dir: Optional[Path] = None, project_root: Optional[Path] = None, logger: Optional[logging.Logger] = None):
+    # Plugin config keys that mean "where this device is". A plugin declaring
+    # any of these in its schema gets the device-wide ``location`` block from
+    # config.json as the *default* for that field, instead of whatever city the
+    # plugin author happened to ship. A value the user set on the plugin itself
+    # always wins -- this only ever replaces the schema default, so an explicit
+    # per-plugin location is still honoured.
+    #
+    # Only these fully-namespaced keys are substituted. A bare ``state`` or
+    # ``city`` key is deliberately left alone: plugins use those for unrelated
+    # things (ledmatrix-elections' ``state`` is a two-letter code, not a place
+    # name), and silently rewriting them would break those plugins.
+    DEVICE_LOCATION_KEYS: Dict[str, str] = {
+        'location_city': 'city',
+        'location_state': 'state',
+        'location_country': 'country',
+    }
+
+    def __init__(self, plugins_dir: Optional[Path] = None, project_root: Optional[Path] = None,
+                 logger: Optional[logging.Logger] = None, config_manager: Optional[Any] = None):
         """
         Initialize the Schema Manager.
         
@@ -34,10 +52,14 @@ class SchemaManager:
             plugins_dir: Base plugins directory path
             project_root: Project root directory path
             logger: Optional logger instance
+            config_manager: Optional config manager, used to resolve the
+                device-wide ``location`` that seeds plugin location defaults.
+                Omitting it simply leaves schema defaults untouched.
         """
         self.logger = logger or logging.getLogger(__name__)
         self.plugins_dir = plugins_dir
         self.project_root = project_root or Path.cwd()
+        self.config_manager = config_manager
         
         # Schema cache: plugin_id -> schema dict
         self._schema_cache: Dict[str, Dict[str, Any]] = {}
@@ -212,10 +234,70 @@ class SchemaManager:
         
         return defaults
     
+    def get_device_location(self) -> Optional[Dict[str, Any]]:
+        """
+        Return the device-wide ``location`` block from config.json, or None.
+
+        This is the City/State/Country the user sets once under General
+        settings. Returns None when there is no config manager wired, the
+        config can't be read, or no location has been configured.
+        """
+        if self.config_manager is None:
+            return None
+        try:
+            config = self.config_manager.load_config()
+        except Exception as e:
+            # A config that can't be read must never stop defaults being
+            # generated -- the plugin's own schema defaults still apply.
+            self.logger.debug(f"Could not read device location from config: {e}")
+            return None
+        if not isinstance(config, dict):
+            return None
+        location = config.get('location')
+        return location if isinstance(location, dict) else None
+
+    def apply_device_location(self, defaults: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        Replace location-shaped schema defaults with the device's own location.
+
+        Without this, a plugin that ships ``"location_city": "Dallas"`` as its
+        schema default silently reports Dallas weather (and centres its radar
+        there) for every user who never opened that plugin's config form --
+        even though they set their real city under General settings. The
+        substituted value is still only a *default*: ``merge_with_defaults``
+        lets any per-plugin value the user saved win over it.
+
+        Mutates and returns ``defaults`` for convenience.
+        """
+        if not defaults:
+            return defaults
+        if not any(key in defaults for key in self.DEVICE_LOCATION_KEYS):
+            return defaults
+
+        location = self.get_device_location()
+        if not location:
+            return defaults
+
+        for key, field in self.DEVICE_LOCATION_KEYS.items():
+            if key not in defaults:
+                continue
+            value = location.get(field)
+            # Only a non-empty string is a real answer; a blank or missing
+            # field means "not configured", which leaves the schema default.
+            if isinstance(value, str) and value.strip():
+                defaults[key] = value.strip()
+
+        return defaults
+
     def generate_default_config(self, plugin_id: str, use_cache: bool = True) -> Dict[str, Any]:
         """
         Generate default configuration for a plugin from its schema.
         
+        Location fields (see ``DEVICE_LOCATION_KEYS``) default to the device's
+        configured location rather than the plugin author's. That substitution
+        is applied on the way out rather than being cached, so changing the
+        device location takes effect without invalidating the defaults cache.
+        
         Args:
             plugin_id: Plugin identifier
             use_cache: If True, return cached defaults if available
@@ -225,7 +307,7 @@ class SchemaManager:
         """
         # Check cache first
         if use_cache and plugin_id in self._defaults_cache:
-            return self._defaults_cache[plugin_id].copy()
+            return self.apply_device_location(self._defaults_cache[plugin_id].copy())
         
         schema = self.load_schema(plugin_id, use_cache=use_cache)
         if not schema:
@@ -249,10 +331,11 @@ class SchemaManager:
         if 'live_priority' not in defaults:
             defaults['live_priority'] = schema.get('properties', {}).get('live_priority', {}).get('default', False)
         
-        # Cache the defaults
+        # Cache the defaults *before* the device location is layered on, so a
+        # later change to the device location is picked up by the next call.
         self._defaults_cache[plugin_id] = defaults.copy()
         
-        return defaults
+        return self.apply_device_location(defaults)
     
     def validate_config_against_schema(self, config: Dict[str, Any], schema: Dict[str, Any], 
                                       plugin_id: Optional[str] = None) -> Tuple[bool, List[str]]:

src/plugin_system/store_manager.py

@@ -1467,8 +1467,11 @@ class PluginStoreManager:
                 # already had. Allowing it costs them a plugin that raises
                 # ModuleNotFoundError at load and is reported only as one line
                 # in the journal. See docs/SPORTS_UNIFICATION.md (phase B4/B6).
-                from src import __version__ as core_version
                 from src.plugin_system import compatibility
+                # On disk, not as imported: this process may predate the core
+                # update that made the plugin compatible. See
+                # compatibility.current_core_version.
+                core_version = compatibility.current_core_version()
 
                 compatible, reason = compatibility.check(manifest, core_version)
                 if not compatible:
@@ -1602,6 +1605,26 @@ class PluginStoreManager:
                     'error': f'Manifest missing required fields: {", ".join(missing_fields)}'
                 }
             
+            # Refuse a plugin that needs a newer core than this one, exactly as
+            # _install_plugin_impl does after its download. Sideloading is an
+            # explicit act rather than an automatic store update, but the floor
+            # is not advice about intent -- it is a statement that the plugin
+            # cannot run here, and letting it through produces the same silent
+            # PluginState.ERROR at load. This was the last of the three routes
+            # in that skipped the check.
+            #
+            # Before the move, so the `finally` below removes the temp tree and
+            # nothing half-installed is left behind.
+            from src.plugin_system import compatibility
+            core_version = compatibility.current_core_version()
+
+            compatible, reason = compatibility.check(manifest, core_version)
+            if not compatible:
+                self.logger.error(
+                    "Refusing to install %s from %s: %s",
+                    plugin_id, repo_url, reason)
+                return {'success': False, 'error': reason}
+
             # Validate version fields consistency (warnings only, not required)
             validation_errors = self._validate_manifest_version_fields(manifest)
             if validation_errors:
@@ -2583,6 +2606,85 @@ class PluginStoreManager:
             self.logger.error(f"Error uninstalling plugin {plugin_id}: {e}")
             return False
     
+    def _gate_pulled_commit(self, plugin_id: str, plugin_path: Path,
+                            previous_sha: Optional[str]) -> bool:
+        """Apply the compatibility gate to a commit that arrived via git pull.
+
+        Every other route into an installed plugin goes through
+        ``install_plugin``, which gates in ``_install_plugin_impl``. This one
+        did not: a ``git pull`` could deliver a manifest flooring above this
+        core and nothing would notice until the plugin failed to load, which
+        surfaces as one line in the journal and a scoreboard that silently
+        stopped appearing.
+
+        Checked after the pull rather than before it, for the same reason
+        ``_install_plugin_impl`` checks after the download: the registry
+        carries no compatibility field, so the incoming floor is only knowable
+        once the new commit is on disk.
+
+        Undone with ``git reset --hard`` rather than by removing the directory.
+        This is a live checkout, the previous commit is still in the object
+        store, and the reset leaves the user on the exact version they were
+        already running -- the same promise ``_reinstall_with_rollback`` makes,
+        reached by the means this path actually has. It is also the gentler
+        option: no window in which the plugin directory does not exist, and no
+        ``.standalone-backup-`` debris if the process dies mid-way.
+
+        A manifest that cannot be read is not evidence of incompatibility, so
+        it allows. ``compatibility.check`` refuses only on evidence for the
+        same reason: a wrong refusal breaks a working install, while a wrong
+        allowance degrades to exactly the behaviour this path had before the
+        gate existed.
+        """
+        manifest_path = plugin_path / "manifest.json"
+        try:
+            with open(manifest_path, 'r', encoding='utf-8') as mf:
+                manifest = json.load(mf)
+        except (OSError, ValueError) as e:
+            self.logger.warning(
+                "Could not read %s after updating %s (%s); allowing the "
+                "update, as an unreadable manifest declares no floor",
+                manifest_path, plugin_id, e)
+            return True
+
+        from src.plugin_system import compatibility
+        core_version = compatibility.current_core_version()
+
+        compatible, reason = compatibility.check(manifest, core_version)
+        if compatible:
+            return True
+
+        self.logger.error("Refusing the update to %s: %s", plugin_id, reason)
+
+        if not previous_sha:
+            self.logger.error(
+                "Cannot roll %s back: the commit it was on before the pull is "
+                "unknown. It is now on a version this core cannot run — "
+                "reinstall it from the plugin store.", plugin_id)
+            return False
+
+        # Safe by construction: update_plugin returns before pulling unless the
+        # tree was clean or successfully stashed, so there are no uncommitted
+        # tracked edits for --hard to discard. The stash is not popped on the
+        # success path either, so the reset leaves the working tree exactly
+        # where a successful pull would have. Say "commit", not "changes".
+        reset = subprocess.run(
+            ['git', '-C', str(plugin_path), 'reset', '--hard', previous_sha],
+            capture_output=True, text=True, timeout=60, check=False)
+        if reset.returncode != 0:
+            self.logger.error(
+                "CRITICAL: could not roll %s back to commit %s: %s. It is left "
+                "on a version this core cannot run; "
+                "`git -C %s reset --hard %s` restores it.",
+                plugin_id, previous_sha[:7],
+                (reset.stderr or reset.stdout or '').strip(),
+                plugin_path, previous_sha)
+        else:
+            self.logger.info(
+                "Rolled %s back to commit %s; it stays on the version it was "
+                "already running.", plugin_id, previous_sha[:7])
+        return False
+
     def _reinstall_with_rollback(self, plugin_id: str, plugin_path: Path) -> bool:
         """Replace an installed plugin with a fresh install, atomically.
 
@@ -2860,6 +2962,8 @@ class PluginStoreManager:
                         status_result = type('obj', (object,), {'stdout': '', 'stderr': 'Status check timed out'})()
                     
                     stash_info = ""
+                    # Whether the pull can be undone without destroying work.
+                    tree_is_recoverable = not has_changes
                     if has_changes:
                         self.logger.info(f"Stashing local changes in {plugin_id} before update")
                         try:
@@ -2873,12 +2977,37 @@ class PluginStoreManager:
                             )
                             if stash_result.returncode == 0:
                                 stash_info = " (local changes were stashed)"
+                                tree_is_recoverable = True
                                 self.logger.info(f"Stashed local changes (including untracked files) for {plugin_id}")
                             else:
                                 self.logger.warning(f"Failed to stash local changes for {plugin_id}: {stash_result.stderr}")
                         except subprocess.TimeoutExpired:
                             self.logger.warning(f"Stash operation timed out for {plugin_id}, proceeding with pull")
 
+                    # Do not pull what cannot be un-pulled.
+                    #
+                    # The compatibility gate below can refuse the commit this
+                    # pull brings down, and its only way back is `git reset
+                    # --hard`, which discards uncommitted tracked edits. Those
+                    # edits are exactly what the stash above exists to protect,
+                    # so a stash that failed or timed out leaves the rollback
+                    # unable to run without destroying them.
+                    #
+                    # A pull does not necessarily refuse on a dirty tree -- git
+                    # merges happily as long as the incoming commit touches
+                    # different files -- so without this the update would
+                    # succeed, the gate would refuse, and the reset would take
+                    # the user's work with it. Refusing here costs an update in
+                    # a case that already went wrong; the alternative costs
+                    # data.
+                    if not tree_is_recoverable:
+                        self.logger.error(
+                            "Refusing to update %s: it has local changes that could "
+                            "not be stashed, and an incompatible update could then "
+                            "only be rolled back by discarding them. Commit or stash "
+                            "them by hand, then update.", plugin_id)
+                        return False
+
                     # Pull from the determined remote branch
                     self.logger.info(f"Pulling from origin/{remote_pull_branch} for {plugin_id}...")
                     pull_result = subprocess.run(
@@ -2901,6 +3030,14 @@ class PluginStoreManager:
                     elif updated_sha:
                         self.logger.info(f"Plugin {plugin_id} updated to commit {updated_sha[:7]}{stash_info}")
 
+                    # The install gate, at the only point on this path where
+                    # it can be answered. Every other route in goes through
+                    # install_plugin, which gates in _install_plugin_impl; this
+                    # one did not, so a pull could deliver a manifest flooring
+                    # above this core and nothing would notice.
+                    if not self._gate_pulled_commit(plugin_id, plugin_path, local_sha):
+                        return False
+
                     self._install_dependencies(plugin_path)
                     return True
 

src/plugin_system/testing/harness.py

@@ -73,6 +73,11 @@ class RenderResult:
     golden_ok: Optional[bool] = None
     golden_diff_pixels: int = 0
     golden_max_delta: int = 0
+    # what display() handed back; the controller skips a mode only on False
+    display_returned: Any = None
+    # empty-frame check: rendered nothing while not reporting "no content"
+    empty_claimed: Optional[bool] = None   # True when that happened
+    empty_ok: Optional[bool] = None        # False only in strict mode
     # fill / scale-up check (populated only for sizes >= 2x the design size)
     fill_checked: bool = False
     fill_ok: Optional[bool] = None       # False only in strict mode
@@ -92,6 +97,8 @@ class RenderResult:
             return False
         if self.fill_ok is False:
             return False
+        if self.empty_ok is False:
+            return False
         return True
 
 
@@ -132,21 +139,25 @@ def _instantiate(plugin_id: str, manifest: Dict[str, Any], plugin_dir: Path,
     return plugin_instance
 
 
-def _render_mode(plugin_instance: Any, mode: str) -> None:
+def _render_mode(plugin_instance: Any, mode: str) -> Any:
     """Render a specific screen. Prefer an explicit display_mode kwarg; otherwise
     drive the plugin's internal mode state machine (first display() call renders
-    modes[current_mode_index] when current_display_mode is None)."""
+    modes[current_mode_index] when current_display_mode is None).
+
+    Returns whatever display() returned. The display controller skips a mode
+    whose display() returns False, so that value decides whether an empty mode
+    is rotated past or sat on -- which makes it worth reporting rather than
+    discarding."""
     sig = inspect.signature(plugin_instance.display)
     if "display_mode" in sig.parameters:
-        plugin_instance.display(force_clear=True, display_mode=mode)
+        return plugin_instance.display(force_clear=True, display_mode=mode)
-        return
 
     modes = getattr(plugin_instance, "modes", None)
     if modes and mode in modes:
         plugin_instance.current_mode_index = list(modes).index(mode)
     if hasattr(plugin_instance, "current_display_mode"):
         plugin_instance.current_display_mode = None
-    plugin_instance.display(force_clear=False)
+    return plugin_instance.display(force_clear=False)
 
 
 def _freeze(freeze_time: Optional[str]):
@@ -234,7 +245,7 @@ def _render_size(plugin_id, manifest, plugin_dir, config, mock_data,
                     logger.warning("update() raised a non-connectivity error for %s [%s]: %s",
                                    plugin_id, mode, e)
             if result.error is None:
-                _render_mode(inst, mode)
+                result.display_returned = _render_mode(inst, mode)
                 result.image = dm.get_image()
                 result.overflow = dm.check_overflow()
         except Exception as e:  # noqa: BLE001 — a display crash is a real failure
@@ -341,6 +352,44 @@ def fill_metrics(image: Image.Image) -> Tuple[float, float, float]:
     return (extent_x, extent_y, ink)
 
 
+def check_empty_claimed(results: List[RenderResult],
+                        strict: bool = False) -> List[RenderResult]:
+    """Flag a mode that rendered nothing without reporting "no content".
+
+    The display controller skips a mode whose ``display()`` returns False, and
+    treats anything else -- including None -- as "content was shown". A mode
+    that draws nothing and does not return False therefore holds whatever is on
+    the panel for its whole display duration. Since a mode switch clears first,
+    that is a blank screen. Two sports plugins shipped exactly this: their
+    ``display()`` returned None on every path, so an out-of-season league sat
+    blank for its full duration rather than being rotated past.
+
+    Warn-only by default, because a blank frame is not automatically wrong: a
+    scroll mode whose first frame is its blank scroll-in buffer renders empty
+    and is behaving correctly. ``strict=True`` sets ``empty_claimed`` such that
+    ``RenderResult.ok`` fails -- opt in per plugin via harness.json
+    ``{"empty_check": "strict"}`` once its modes are known to draw on the
+    fixture data.
+
+    Note this can only catch what the fixtures actually render. A plugin whose
+    harness fixture seeds content never exercises its empty path here; the
+    source-level gate in the plugins repo covers that case.
+    """
+    for r in results:
+        if r.image is None or r.error is not None:
+            continue
+        # An explicit False is the plugin correctly saying "nothing to show".
+        if r.display_returned is False:
+            continue
+        if r.image.convert("L").point(
+                lambda p: 255 if p > _LIT_THRESHOLD else 0).getbbox() is not None:
+            continue
+        r.empty_claimed = True
+        if strict:
+            r.empty_ok = False
+    return results
+
+
 def check_scale_up(results: List[RenderResult],
                    design_size: Tuple[int, int] = (128, 32),
                    min_extent: float = _MIN_FILL_EXTENT,

src/plugin_system/testing/visual_display_manager.py

@@ -70,6 +70,7 @@ class VisualTestDisplayManager:
         # Canvas
         self.image = Image.new('RGB', (width, height), (0, 0, 0))
         self.draw = ImageDraw.Draw(self.image)
+        self.draw.fontmode = "1"  # Match production: 1-bit text, so goldens show what the panel shows.
 
         # Matrix proxy (plugins access display_manager.matrix.width/height)
         self.matrix = _MatrixProxy(width, height)
@@ -184,6 +185,7 @@ class VisualTestDisplayManager:
         self.clear_called = True
         self.image = Image.new('RGB', (self._width, self._height), (0, 0, 0))
         self.draw = ImageDraw.Draw(self.image)
+        self.draw.fontmode = "1"  # Match production: 1-bit text, so goldens show what the panel shows.
 
     def update_display(self):
         """No-op for hardware; marks that display was updated."""
@@ -211,6 +213,7 @@ class VisualTestDisplayManager:
             self.matrix = _MatrixProxy(target_w, target_h)
             self.image = Image.new('RGB', (target_w, target_h), (0, 0, 0))
             self.draw = ImageDraw.Draw(self.image)
+            self.draw.fontmode = "1"  # Match production: 1-bit text, so goldens show what the panel shows.
             yield
         finally:
             self._width, self._height = prev_w, prev_h
@@ -569,6 +572,7 @@ class VisualTestDisplayManager:
         self.draw_calls = []
         self.image = Image.new('RGB', (self._width, self._height), (0, 0, 0))
         self.draw = ImageDraw.Draw(self.image)
+        self.draw.fontmode = "1"  # Match production: 1-bit text, so goldens show what the panel shows.
         self._scrolling_state = {
             'is_scrolling': False,
             'last_scroll_activity': 0,
@@ -582,3 +586,4 @@ class VisualTestDisplayManager:
         """Clean up resources."""
         self.image = Image.new('RGB', (self._width, self._height), (0, 0, 0))
         self.draw = ImageDraw.Draw(self.image)
+        self.draw.fontmode = "1"  # Match production: 1-bit text, so goldens show what the panel shows.

src/startup_validator.py

@@ -15,16 +15,23 @@ from src.logging_config import get_logger
 class StartupValidator:
     """Validates system state on startup."""
     
-    def __init__(self, config_manager: Any, plugin_manager: Optional[Any] = None) -> None:
+    def __init__(self, config_manager: Any, plugin_manager: Optional[Any] = None,
+                 cache_manager: Optional[Any] = None) -> None:
         """
         Initialize the startup validator.
 
         Args:
             config_manager: ConfigManager instance
             plugin_manager: Optional PluginManager instance
+            cache_manager: The CacheManager the application will actually use.
+                Pass it. Without one this validator builds its own just to read
+                a directory path, which reports on a cache the app does not
+                use and leaves behind a cleanup thread that nothing stops --
+                validation runs twice per startup, so that was two of them.
         """
         self.config_manager = config_manager
         self.plugin_manager = plugin_manager
+        self.cache_manager = cache_manager
         self.logger = get_logger(__name__)
         self.errors: List[str] = []
         self.warnings: List[str] = []
@@ -38,6 +45,11 @@ class StartupValidator:
         """
         self.logger.info("Starting startup validation...")
 
+        # Fresh lists each run — without this, calling validate_all() twice
+        # duplicated every message.
+        self.errors = []
+        self.warnings = []
+
         # Validate configuration
         self._validate_config()
         
@@ -51,6 +63,9 @@ class StartupValidator:
         if self.plugin_manager:
             self._validate_plugins()
 
+        # Warn when the running systemd unit no longer matches the repo's
+        self._validate_systemd_units()
+        
         is_valid = len(self.errors) == 0
         
         if is_valid:
@@ -62,6 +77,80 @@ class StartupValidator:
         
         return (is_valid, self.errors.copy(), self.warnings.copy())
     
+    #: Units this project installs, and where each is installed to.
+    _UNITS = (
+        ("systemd/ledmatrix.service", "/etc/systemd/system/ledmatrix.service"),
+        ("systemd/ledmatrix-web.service", "/etc/systemd/system/ledmatrix-web.service"),
+    )
+
+    def _validate_systemd_units(self) -> None:
+        """Warn when an installed unit has drifted from the repo's template.
+
+        Nothing re-applies these after the first install. `git pull` -- which is
+        what the web UI's update button runs -- brings a new template into the
+        checkout, but nothing copies it to /etc/systemd/system and nothing runs
+        `systemctl daemon-reload`, so the unit that actually runs is whatever
+        first_time_install.sh wrote on day one.
+
+        That makes every hardening added to a unit inert on existing installs.
+        Measured on one rig: the installed unit was thirteen days older than the
+        repo's and differed in content, so a MemoryMax the repo had specified
+        was not being enforced at all -- `systemctl show` reported
+        MemoryMax=infinity.
+
+        A warning rather than an error, and certainly not a silent rewrite:
+        editing files under /etc and restarting services is the installer's job,
+        not something a display process should do to a machine while it boots.
+        The remedy is to re-run scripts/install/install_service.sh.
+        """
+        try:
+            project_root = Path(__file__).resolve().parent.parent
+            for template_rel, installed_path in self._UNITS:
+                template = project_root / template_rel
+                installed = Path(installed_path)
+                if not template.is_file() or not installed.is_file():
+                    continue
+
+                # The template carries placeholders the installer substitutes,
+                # so compare the substituted form rather than the raw file.
+                expected = template.read_text(encoding="utf-8")
+                expected = expected.replace("__PROJECT_ROOT_DIR__", str(project_root))
+                expected = expected.replace("__USER__", "root")
+
+                try:
+                    actual = installed.read_text(encoding="utf-8")
+                except PermissionError:
+                    continue
+
+                if self._unit_body(expected) != self._unit_body(actual):
+                    self.warnings.append(
+                        f"{installed.name} differs from {template_rel}; the "
+                        "installed unit is not refreshed by an update, so "
+                        "settings added to the template are not in effect. "
+                        "Re-run scripts/install/install_service.sh to apply them."
+                    )
+        except OSError as e:
+            self.logger.debug("Could not compare systemd units: %s", e)
+
+    @staticmethod
+    def _unit_body(text: str) -> str:
+        """A unit's meaningful lines, in order: no comments, no blanks.
+
+        Order is preserved deliberately. This used to sort, which made the
+        comparison insensitive to two changes that matter in a systemd unit:
+        repeated directives such as ExecStartPre= and ExecStartPost= run in
+        the order they appear, and a directive that moves between [Unit],
+        [Service] and [Install] means something different -- or nothing --
+        where it lands. A drift check that normalises those away reports no
+        drift for a unit that has genuinely changed.
+        """
+        lines = []
+        for line in text.splitlines():
+            line = line.strip()
+            if line and not line.startswith("#"):
+                lines.append(line)
+        return "\n".join(lines)
+
     def _validate_config(self) -> None:
         """Validate configuration files."""
         try:
@@ -86,8 +175,20 @@ class StartupValidator:
     def _validate_cache_directory(self) -> None:
         """Validate cache directory permissions."""
         try:
+            cache_manager = self.cache_manager
+            if cache_manager is None:
+                # No caller supplied one (older embedders, direct use in a
+                # script). Build one, but do not leave its cleanup thread
+                # running behind us -- this instance is discarded on the next
+                # line but the thread is a closure over it, so it would never
+                # be collected.
                 from src.cache_manager import CacheManager
                 cache_manager = CacheManager()
+                try:
+                    cache_dir = cache_manager.get_cache_dir()
+                finally:
+                    cache_manager.stop_cleanup_thread()
+            else:
                 cache_dir = cache_manager.get_cache_dir()
             
             if not cache_dir:

src/vegas_mode/config.py

@@ -104,15 +104,53 @@ class VegasModeConfig:
     overflow_mode: str = "rotate"
 
     # Cap on one plugin's share of a cycle, as a multiple of display width.
-    # A single ticker returning 7,000px would otherwise hold the panel for over
+    # 0 (the default) disables the cap, so every plugin contributes all of its
-    # two minutes. Overflow is deferred to later cycles rather than discarded.
+    # content and is always entered at its beginning.
-    # 0 disables the cap.
+    #
-    max_plugin_width_ratio: float = 3.0
+    # Capping was the default until it proved to cost more than it bought.
+    # Measured over a 17-plugin fleet on a 512px panel, only four plugins were
+    # ever wide enough to hit a 3.0 cap; for those four it produced two visible
+    # faults. Content resumed mid-item on each appearance (a news ticker entered
+    # at column 6027 of its own strip), and the final window of a rotation was
+    # whatever happened to be left — 348px of a 1840px stocks ticker, seven
+    # seconds of panel time. Both read as the display being broken rather than
+    # as deferral working.
+    #
+    # A wide plugin does hold the panel for a long time uncapped: set the cap
+    # per plugin with vegas_max_width_screens where that matters, rather than
+    # globally where it mostly hurts plugins that were never the problem.
+    max_plugin_width_ratio: float = 0.0
 
     # Plugin management
     plugin_order: List[str] = field(default_factory=list)
     excluded_plugins: Set[str] = field(default_factory=set)
 
+    # --- Live content in the ticker -------------------------------------
+    #
+    # By default a live game preempts Vegas entirely: the display controller
+    # refuses to run the ticker while any plugin reports live priority, and you
+    # get the full-screen scoreboard instead. Set live_in_ticker to keep the
+    # marquee running and let live content take extra turns within it.
+    #
+    # The rotation is otherwise a strict round robin -- every plugin appears
+    # exactly once per cycle -- so with a dozen plugins enabled a live score
+    # comes round once a lap and can be minutes old on screen. Weighting lets a
+    # plugin claim several slots per cycle instead.
+    #
+    # Weights are per plugin, not per game: a scoreboard showing four live
+    # games still occupies one slot at a time, and rotates its own games within
+    # that slot using its own favorite_live_boost.
+    live_in_ticker: bool = False
+
+    # Slots per cycle for a plugin reporting live content. 1 disables the boost
+    # and restores the plain round robin.
+    live_weight: int = 3
+
+    # Slots per cycle for a plugin whose live content involves a favorite team.
+    # Only plugins implementing get_vegas_priority_weight() can claim this --
+    # the core cannot tell whose game is on, so the plugin reports it.
+    favorite_live_weight: int = 5
+
     # Performance settings
     target_fps: int = 125  # Target frame rate
     buffer_ahead: int = 2  # Number of plugins to buffer ahead
@@ -159,10 +197,16 @@ class VegasModeConfig:
             lead_in_width=int(vegas_config.get('lead_in_width', 0)),
             plugins_per_cycle=int(vegas_config.get('plugins_per_cycle', 6)),
             max_plugin_width_ratio=float(
-                vegas_config.get('max_plugin_width_ratio', 3.0)),
+                vegas_config.get('max_plugin_width_ratio', 0.0)),
             overflow_mode=str(vegas_config.get('overflow_mode', 'rotate')),
             plugin_order=list(vegas_config.get('plugin_order', [])),
             excluded_plugins=set(vegas_config.get('excluded_plugins', [])),
+            live_in_ticker=bool(vegas_config.get('live_in_ticker', False)),
+            # Clamped: a weight below 1 would drop the plugin from the rotation
+            # entirely, and a very large one starves everything else.
+            live_weight=max(1, min(10, int(vegas_config.get('live_weight', 3)))),
+            favorite_live_weight=max(
+                1, min(10, int(vegas_config.get('favorite_live_weight', 5)))),
             target_fps=int(vegas_config.get('target_fps', 125)),
             buffer_ahead=int(vegas_config.get('buffer_ahead', 2)),
             frame_based_scrolling=vegas_config.get('frame_based_scrolling', True),
@@ -192,6 +236,9 @@ class VegasModeConfig:
             'lead_in_width': self.lead_in_width,
             'plugins_per_cycle': self.plugins_per_cycle,
             'max_plugin_width_ratio': self.max_plugin_width_ratio,
+            'live_in_ticker': self.live_in_ticker,
+            'live_weight': self.live_weight,
+            'favorite_live_weight': self.favorite_live_weight,
             'overflow_mode': self.overflow_mode,
             'plugin_order': self.plugin_order,
             'excluded_plugins': list(self.excluded_plugins),
@@ -359,6 +406,15 @@ class VegasModeConfig:
 
         if 'enabled' in vegas_config:
             self.enabled = vegas_config['enabled']
+        if 'live_in_ticker' in vegas_config:
+            self.live_in_ticker = bool(vegas_config['live_in_ticker'])
+        # Clamped exactly as from_config does: a weight below 1 would drop the
+        # plugin from the rotation, and a huge one starves everything else.
+        if 'live_weight' in vegas_config:
+            self.live_weight = max(1, min(10, int(vegas_config['live_weight'])))
+        if 'favorite_live_weight' in vegas_config:
+            self.favorite_live_weight = max(
+                1, min(10, int(vegas_config['favorite_live_weight'])))
         if 'scroll_speed' in vegas_config:
             self.scroll_speed = float(vegas_config['scroll_speed'])
         if 'separator_width' in vegas_config:

src/vegas_mode/coordinator.py

@@ -12,6 +12,7 @@ Supports three display modes per plugin:
 """
 
 import logging
+import math
 import time
 import threading
 from typing import Optional, Dict, Any, List, Callable, TYPE_CHECKING
@@ -30,6 +31,33 @@ if TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 
+#: Degradation threshold, as a fraction of target_fps. A marquee jitters a
+#: little all the time, so "anything under target" would report constantly and
+#: mean nothing; 90% of target is the point where a shortfall is real. At a
+#: 60fps target that is 54fps -- 55fps is a normal wobble and stays at DEBUG,
+#: which is deliberate, not an off-by-one.
+_FPS_HEALTHY_FRACTION = 0.9
+
+#: A healthy marquee still reports this often, so silence means stopped
+#: rather than fine.
+_FPS_HEARTBEAT_INTERVAL = 300.0
+
+
+def _percentile(ordered: List[float], fraction: float) -> float:
+    """Nearest-rank percentile of an already-sorted list.
+
+    Index ceil(n * fraction) - 1, so 100 samples at 0.99 give the 99th-ranked
+    value. The obvious int(n * fraction) is off by one and, at exactly 100
+    samples, lands on the maximum -- which is the number already reported
+    alongside this one as the worst frame, so the two columns would agree
+    precisely when the sample was smallest.
+    """
+    if not ordered:
+        return 0.0
+    index = math.ceil(len(ordered) * fraction) - 1
+    return ordered[min(len(ordered) - 1, max(0, index))]
+
+
 class VegasModeCoordinator:
     """
     Orchestrates Vegas scroll mode operation.
@@ -80,6 +108,11 @@ class VegasModeCoordinator:
         self._is_active = False
         self._is_paused = False
         self._should_stop = False
+        # Frame-rate health, tracked across run_iteration() calls so the
+        # heartbeat is one-per-interval rather than one-per-cycle, and so a
+        # recovery spanning two cycles is still reported. Reset on start().
+        self._fps_last_health_log = 0.0
+        self._fps_was_degraded = False
         self._state_lock = threading.Lock()
 
         # Live priority tracking
@@ -232,6 +265,11 @@ class VegasModeCoordinator:
             self._is_active = True
             self._should_stop = False
             self._start_time = time.time()
+            # A fresh run starts with a clean health slate: no stale
+            # "was degraded" from the previous run, and a heartbeat that is
+            # due immediately so the first sample confirms the marquee is up.
+            self._fps_last_health_log = 0.0
+            self._fps_was_degraded = False
 
         # Line up the next group immediately, so the first extension is already
         # warm rather than stalling the scroll to fetch it.
@@ -379,14 +417,36 @@ class VegasModeCoordinator:
             duration = self.render_pipeline.get_dynamic_duration()
         start_time = time.time()
         frame_count = 0
-        fps_log_interval = 5.0  # Log FPS every 5 seconds
+        fps_log_interval = 5.0  # Sample FPS every 5 seconds
-        last_fps_log_time = start_time
+        # Health state lives on the coordinator, not here: run_iteration() is
+        # called once per cycle, so locals reset every few seconds. That made
+        # `last_fps_health_log = 0.0` fire the "heartbeat" on the first sample
+        # of every iteration rather than once per interval, and a recovery
+        # that crossed an iteration boundary was never reported at all --
+        # was_degraded had already gone back to False.
+        # Monotonic, and deliberately not start_time: start_time is wall
+        # clock and is used below to report the iteration's duration. Mixing
+        # the two here would make every delta hugely negative and silence the
+        # frame-rate reporting altogether.
+        last_fps_log_time = time.monotonic()
         fps_frame_count = 0
+        # A mean hides stutter completely. At 120fps a five-second window is
+        # ~600 frames, so a 200ms freeze -- plainly visible on a marquee --
+        # moves the average from 120.0 to 115.4 and reads as healthy. What a
+        # viewer actually notices is the worst frame, so track that too.
+        frame_worst = 0.0
+        frame_times: List[float] = []
 
         logger.info("Starting Vegas iteration for %.1fs", duration)
 
         while True:
-            frame_started = time.time()
+            # Monotonic, like the FPS window below. These devices have no RTC,
+            # so the wall clock jumps by however wrong boot time was the moment
+            # NTP first syncs. A backward jump makes frame_elapsed negative,
+            # and `frame_interval - frame_elapsed` then sleeps for longer than
+            # the whole budget -- the render loop stalls for the size of the
+            # correction. A forward jump inflates p99 and worst-frame instead.
+            frame_started = time.monotonic()
 
             # Check for STATIC mode plugin that should pause scroll
             static_plugin = self._check_static_plugin_trigger()
@@ -414,23 +474,58 @@ class VegasModeCoordinator:
             # quarter of the budget spent not rendering. Subtracting the work
             # already done keeps the pacing target while reclaiming that time,
             # and yields the GIL either way so other threads still run.
-            frame_elapsed = time.time() - frame_started
+            frame_elapsed = time.monotonic() - frame_started
             time.sleep(max(0.0, frame_interval - frame_elapsed))
 
+            # Measured before the sleep: time spent working, not pacing.
+            if frame_elapsed > frame_worst:
+                frame_worst = frame_elapsed
+            frame_times.append(frame_elapsed)
+
             # Increment frame count and check for interrupt periodically
             frame_count += 1
             fps_frame_count += 1
 
-            # Periodic FPS logging
+            # Periodic FPS logging. Reported at INFO only when the frame rate
-            current_time = time.time()
+            # is actually worth an operator's attention -- a shortfall against
+            # target, or the recovery from one -- with a slow heartbeat so a
+            # healthy marquee still shows a pulse.
+            #
+            # Measured over two hours on a running rig: 1410 samples, 98.5%
+            # of them within 10% of target. The 1.5% that were not included a
+            # reading of 8.6fps against a target of 60 -- a real stall, and
+            # completely invisible inside 1389 lines reading "59.6".
+            # Monotonic: every use of this value in the block below is a
+            # duration, and these devices have no RTC, so the wall clock jumps
+            # by however wrong boot time was the moment NTP first syncs. That
+            # would not only mis-fire the heartbeat, it would corrupt the
+            # frame rate itself, since fps is frames divided by this delta.
+            current_time = time.monotonic()
             if current_time - last_fps_log_time >= fps_log_interval:
                 fps = fps_frame_count / (current_time - last_fps_log_time)
+                p99 = _percentile(sorted(frame_times), 0.99)
+                target = self.vegas_config.target_fps
+                degraded = target > 0 and fps < target * _FPS_HEALTHY_FRACTION
+                due = (current_time - self._fps_last_health_log
+                       >= _FPS_HEARTBEAT_INTERVAL)
+                if degraded or self._fps_was_degraded or due:
                     logger.info(
-                    "Vegas FPS: %.1f (target: %d, frames: %d)",
+                        "Vegas FPS: %.1f (target: %d, frames: %d) p99 %.1fms worst %.1fms",
-                    fps, self.vegas_config.target_fps, fps_frame_count
+                        fps, target, fps_frame_count,
+                        p99 * 1000.0, frame_worst * 1000.0
                     )
+                    self._fps_last_health_log = current_time
+                else:
+                    logger.debug(
+                        "Vegas FPS: %.1f (target: %d, frames: %d) p99 %.1fms worst %.1fms",
+                        fps, target, fps_frame_count,
+                        p99 * 1000.0, frame_worst * 1000.0
+                    )
+                self._fps_was_degraded = degraded
                 last_fps_log_time = current_time
                 fps_frame_count = 0
+                frame_worst = 0.0
+                frame_times.clear()
 
             if (self._interrupt_check and
                     frame_count % self._interrupt_check_interval == 0):
@@ -497,6 +592,12 @@ class VegasModeCoordinator:
         if not self._live_priority_check:
             return False
 
+        if self.vegas_config.live_in_ticker:
+            # The ticker keeps live content rather than yielding to it; the
+            # extra turns are arranged in the rotation itself, so there is
+            # nothing to pause for.
+            return False
+
         try:
             live_mode = self._live_priority_check()
             if live_mode:

src/vegas_mode/plugin_adapter.py

@@ -68,7 +68,22 @@ class PluginAdapter:
         # always the same opening items.
         self._item_offsets: dict = {}
 
-        logger.info(
+        # What the matching entry in _item_offsets is an offset *into*, as
+        # (kind, size). An offset only means anything against the content it
+        # was derived from, and there are three incompatible kinds:
+        #
+        #   ('rows', n)  index into a list of n images
+        #   ('cuts', n)  index into the n item boundaries of one image
+        #   ('cols', w)  pixel column in a w-wide image with no item boundaries
+        #
+        # Without this the offsets were reused across kinds — a plugin that
+        # returned one wide image on one fetch and several rows on the next had
+        # a pixel column of 1400 read back as a row index — and across content
+        # changes, where a column recorded against a 9,793px news strip pointed
+        # into unrelated headlines once the strip refreshed to 9,505px.
+        self._offset_shapes: dict = {}
+
+        logger.debug(
             "PluginAdapter initialized: display=%dx%d",
             self.display_width, self.display_height
         )
@@ -94,7 +109,7 @@ class PluginAdapter:
         Returns:
             List of PIL Images representing plugin content, or None if no content
         """
-        logger.info(
+        logger.debug(
             "[%s] Getting content (class=%s)",
             plugin_id, plugin.__class__.__name__
         )
@@ -103,7 +118,7 @@ class PluginAdapter:
         cached = self._get_cached(plugin_id)
         if cached is not None:
             total_width = sum(img.width for img in cached)
-            logger.info(
+            logger.debug(
                 "[%s] Using cached content: %d images, %dpx total",
                 plugin_id, len(cached), total_width
             )
@@ -111,46 +126,46 @@ class PluginAdapter:
 
         # Try native Vegas content method first
         has_native = hasattr(plugin, 'get_vegas_content')
-        logger.info("[%s] Has get_vegas_content: %s", plugin_id, has_native)
+        logger.debug("[%s] Has get_vegas_content: %s", plugin_id, has_native)
         if has_native:
             content = self._get_native_content(plugin, plugin_id, offscreen_only)
             if content:
                 total_width = sum(img.width for img in content)
-                logger.info(
+                logger.debug(
                     "[%s] Native content SUCCESS: %d images, %dpx total",
                     plugin_id, len(content), total_width
                 )
                 return self._finalize(content, plugin_id, 'native', plugin)
-            logger.info("[%s] Native content returned None", plugin_id)
+            logger.debug("[%s] Native content returned None", plugin_id)
 
         # Try to get scroll_helper's cached image (for scrolling plugins like stocks/odds)
         has_scroll_helper = hasattr(plugin, 'scroll_helper')
-        logger.info("[%s] Has scroll_helper: %s", plugin_id, has_scroll_helper)
+        logger.debug("[%s] Has scroll_helper: %s", plugin_id, has_scroll_helper)
         content = self._get_scroll_helper_content(plugin, plugin_id, offscreen_only)
         if content:
             total_width = sum(img.width for img in content)
-            logger.info(
+            logger.debug(
                 "[%s] ScrollHelper content SUCCESS: %d images, %dpx total",
                 plugin_id, len(content), total_width
             )
             return self._finalize(content, plugin_id, 'scroll_helper', plugin)
         if has_scroll_helper:
-            logger.info("[%s] ScrollHelper content returned None", plugin_id)
+            logger.debug("[%s] ScrollHelper content returned None", plugin_id)
 
         if offscreen_only:
             # Display capture needs the shared canvas; leave it to the caller.
-            logger.info(
+            logger.debug(
                 "[%s] Needs display capture, deferring to the render thread",
                 plugin_id
             )
             return None
 
         # Fall back to display capture
-        logger.info("[%s] Trying fallback display capture...", plugin_id)
+        logger.debug("[%s] Trying fallback display capture...", plugin_id)
         content = self._capture_display_content(plugin, plugin_id)
         if content:
             total_width = sum(img.width for img in content)
-            logger.info(
+            logger.debug(
                 "[%s] Fallback capture SUCCESS: %d images, %dpx total",
                 plugin_id, len(content), total_width
             )
@@ -211,7 +226,7 @@ class PluginAdapter:
             kept.append(result.image)
 
         if not kept:
-            logger.info(
+            logger.debug(
                 "[%s] All %d image(s) from %s were blank — contributing nothing",
                 plugin_id, len(images), source
             )
@@ -220,14 +235,14 @@ class PluginAdapter:
         trimmed_width = sum(img.width for img in kept)
 
         if trimmed_width < self.config.min_plugin_width:
-            logger.info(
+            logger.debug(
                 "[%s] Trimmed content %dpx is below min_plugin_width %dpx — skipping",
                 plugin_id, trimmed_width, self.config.min_plugin_width
             )
             return None
 
         if trimmed_width != original_width or dropped_blank:
-            logger.info(
+            logger.debug(
                 "[%s] Trimmed %s content: %dpx -> %dpx (%.0f%% reclaimed), "
                 "%d image(s) kept, %d blank dropped",
                 plugin_id, source, original_width, trimmed_width,
@@ -398,6 +413,88 @@ class PluginAdapter:
             return 0
         return int(self.display_width * ratio)
 
+    def _resume_offset(self, plugin_id: str, shape: Tuple[str, int]) -> int:
+        """
+        The plugin's stored rotation offset, if it still applies.
+
+        An offset is only meaningful against content shaped the way it was
+        when the offset was recorded. When the shape has changed — a different
+        number of rows, a re-rendered strip with different item boundaries —
+        the stored value points somewhere arbitrary, so rotation restarts.
+
+        Args:
+            plugin_id: Plugin identifier
+            shape: (kind, size) describing what an offset would index into now
+
+        Returns:
+            The stored offset, or 0 when it no longer applies
+        """
+        if self._offset_shapes.get(plugin_id) != shape:
+            if plugin_id in self._item_offsets:
+                logger.debug(
+                    "[%s] Content is %s now, was %s — restarting the rotation "
+                    "rather than resuming at a position that no longer means "
+                    "anything", plugin_id, shape,
+                    self._offset_shapes.get(plugin_id))
+            self._item_offsets.pop(plugin_id, None)
+            self._offset_shapes[plugin_id] = shape
+            return 0
+        return self._item_offsets.get(plugin_id, 0)
+
+    def _record_offset(
+        self, plugin_id: str, offset: int, shape: Tuple[str, int]
+    ) -> None:
+        """Store where the next window should resume, with what it indexes."""
+        if offset:
+            self._item_offsets[plugin_id] = offset
+            self._offset_shapes[plugin_id] = shape
+        else:
+            # A wrapped-to-zero rotation is the same as no state at all, and
+            # keeping the key would report a window as active when the next
+            # pass starts from the top anyway.
+            self._item_offsets.pop(plugin_id, None)
+            self._offset_shapes.pop(plugin_id, None)
+
+    def _clear_offset(self, plugin_id: str) -> None:
+        """Forget any rotation state for a plugin."""
+        self._item_offsets.pop(plugin_id, None)
+        self._offset_shapes.pop(plugin_id, None)
+
+    def _merge_trailing_runt(self, end: int, width: int, budget: int) -> int:
+        """
+        Extend a window to the end of the content when what would be left over
+        is too small to be worth its own pass.
+
+        Windows were placed by walking forward from the last one, which makes
+        the final window whatever happens to remain. Measured on a live panel
+        that produced a 1,840px stocks ticker splitting 1,492 + 348 — the
+        second pass showing seven seconds of content before cutting, which
+        reads as the display failing rather than as a rotation.
+
+        Absorbing the remainder overruns the budget by less than one window
+        floor, which is a better trade than a fragment: the budget is a guard
+        against one plugin holding the panel for minutes, not a hard limit.
+
+        Args:
+            end: Column the window would otherwise end at
+            width: Full content width
+            budget: Width budget being applied
+
+        Returns:
+            ``end``, or ``width`` when the remainder is below the floor
+        """
+        remainder = width - end
+        # Measured against the budget rather than the panel: snapping to item
+        # boundaries means an ordinary window already lands short of the budget
+        # (a 512px budget over 182px-pitch items yields 348px windows), so an
+        # absolute floor would merge windows that were never fragments. Half a
+        # budget separates "a short last pass" from "a sliver", and caps the
+        # overrun this can cause at 1.5 budgets.
+        floor = budget // 2
+        if 0 < remainder < floor:
+            return width
+        return end
+
     def _apply_width_budget(
         self, images: List[Image.Image], plugin_id: str,
         plugin: Optional['BasePlugin'] = None
@@ -435,45 +532,62 @@ class PluginAdapter:
 
         if not budget or total <= budget:
             # Fits, so reset rotation — the whole segment is being shown.
-            self._item_offsets.pop(plugin_id, None)
+            self._clear_offset(plugin_id)
             return images
 
         if len(images) == 1:
             return [self._crop_to_budget(images[0], budget, plugin_id, mode)]
 
+        shape = ('rows', len(images))
         if mode == 'truncate':
             # Ordered content: always show from the top. Deliberately does not
             # advance the offset, so the same opening items appear every time
             # rather than the viewer being shown the middle of a ranked list.
             start = 0
         else:
-            start = self._item_offsets.get(plugin_id, 0) % len(images)
+            start = self._resume_offset(plugin_id, shape) % len(images)
         selected: List[Image.Image] = []
         used = 0
         consumed = 0
 
         # Walk forward from the rotation offset, taking whole items only, so a
         # cut never lands in the middle of one.
+        #
+        # A window may overrun the budget while it is still shorter than the
+        # runt floor, for the same reason _merge_trailing_runt exists on the
+        # single-image path: a pass far shorter than its neighbours reads as
+        # the display failing rather than as a rotation. Rows of 450, 450 and
+        # 100 against a 512px budget used to give the 100 a pass of its own --
+        # two seconds against nine. Wrapping does not prevent that, because it
+        # only helps when the row wrapped to actually fits.
+        floor = budget // 2
         for step in range(len(images)):
             img = images[(start + step) % len(images)]
             cost = img.width
             if selected:
                 cost += self._row_gap(selected[-1], img)
             if selected and used + cost > budget:
+                # Keep the overrun bounded at the same 1.5 budgets the
+                # single-image path allows. A next row too wide to absorb
+                # leaves a short window standing -- better than a window of
+                # 1.9 budgets, and the same trade the always-take-the-first
+                # rule below already makes.
+                if used >= floor or used + cost > budget + floor:
                     break
             selected.append(img)
             used += cost
             consumed += 1
 
         if mode == 'truncate':
-            logger.info(
+            logger.debug(
                 "[%s] Width budget %dpx: showing the first %d of %d row(s) "
                 "(%dpx incl. gaps); the rest are not shown (overflow=truncate)",
                 plugin_id, budget, len(selected), len(images), used
             )
         else:
-            self._item_offsets[plugin_id] = (start + consumed) % len(images)
+            self._record_offset(
-            logger.info(
+                plugin_id, (start + consumed) % len(images), shape)
+            logger.debug(
                 "[%s] Width budget %dpx: showing %d of %d row(s) (%dpx incl. gaps) "
                 "from offset %d; remainder deferred to a later cycle",
                 plugin_id, budget, len(selected), len(images), used, start
@@ -490,16 +604,13 @@ class PluginAdapter:
 
         The cut is snapped to the nearest blank column so it does not slice
         through a glyph or logo and leave half a character at the panel edge.
-        """
-        if mode == 'truncate':
-            # Always the start of the strip, so a ranked table is never entered
-            # from the middle.
-            offset = 0
-        else:
-            offset = self._item_offsets.get(plugin_id, 0)
-            if offset >= img.width:
-                offset = 0
 
+        Rotation is tracked as an index into the strip's item boundaries rather
+        than as a pixel column, because a ticker re-renders between fetches. A
+        column recorded against one render points at unrelated content in the
+        next as soon as anything ahead of it changes width — a digit in a
+        price, a shorter headline. The Nth boundary stays the Nth boundary.
+        """
         # Cut only where the plugin left a real gap between items. Snapping to
         # any blank column used to pick the single-column gaps between
         # characters, splitting a word and orphaning its tail into the next
@@ -514,10 +625,18 @@ class PluginAdapter:
             # budget exactly. The gap rule exists to protect discrete items
             # (words, ticker entries); it would be wrong to let a solid image
             # escape the cap in its name.
-            end = min(offset + budget, img.width)
+            #
+            # With no items to index, the offset here has to stay a column, so
+            # it is only reusable while the image keeps its width.
+            shape = ('cols', img.width)
+            offset = 0 if mode == 'truncate' else self._resume_offset(
+                plugin_id, shape)
+            end = self._merge_trailing_runt(
+                min(offset + budget, img.width), img.width, budget)
             if mode != 'truncate':
-                self._item_offsets[plugin_id] = 0 if end >= img.width else end
+                self._record_offset(
-            logger.info(
+                    plugin_id, 0 if end >= img.width else end, shape)
+            logger.debug(
                 "[%s] Width budget %dpx: cropped continuous %dpx image to "
                 "[%d:%d] (no item gaps of %dpx+ to align to)%s",
                 plugin_id, budget, img.width, offset, end, min_run,
@@ -528,8 +647,15 @@ class PluginAdapter:
         # Cut mid-gap so the content either side keeps some breathing room.
         cuts = sorted({0, img.width} | {(a + b) // 2 for a, b in gaps})
 
-        start = max((c for c in cuts if c <= offset), default=0)
+        shape = ('cuts', len(cuts))
-        later = [c for c in cuts if c > start]
+        index = 0 if mode == 'truncate' else self._resume_offset(
+            plugin_id, shape)
+        # Clamped rather than wrapped: a stale index past the end means the
+        # strip shrank, and restarting reads better than landing near the end.
+        start_index = index if 0 <= index < len(cuts) - 1 else 0
+        start = cuts[start_index]
+
+        later = cuts[start_index + 1:]
         if not later:
             end = img.width
         else:
@@ -537,15 +663,22 @@ class PluginAdapter:
             # No boundary inside the budget: take the next one and overrun,
             # because the alternative is cutting through an item.
             end = max(within) if within else min(later)
+        end = self._merge_trailing_runt(end, img.width, budget)
+        # Every candidate for `end` came from `cuts` (which includes img.width),
+        # so this always resolves; the fallback is defensive only.
+        end_index = cuts.index(end) if end in cuts else len(cuts) - 1
 
         if mode != 'truncate':
-            # Next cycle resumes where this one stopped; wrap when the strip ends.
+            # Next cycle resumes at the boundary this one stopped on; wrap when
-            self._item_offsets[plugin_id] = 0 if end >= img.width else end
+            # the strip ends.
+            self._record_offset(
+                plugin_id, 0 if end >= img.width else end_index, shape)
 
-        logger.info(
+        logger.debug(
             "[%s] Width budget %dpx: cropped single %dpx image to [%d:%d] "
-            "(%dpx) at item boundaries, %s",
+            "(%dpx) at item boundaries %d-%d of %d, %s",
             plugin_id, budget, img.width, start, end, end - start,
+            start_index, end_index, len(cuts) - 1,
             "showing the start only (overflow=truncate)"
             if mode == 'truncate' else "window advances next cycle"
         )
@@ -565,7 +698,7 @@ class PluginAdapter:
             List of images or None
         """
         try:
-            logger.info("[%s] Native: calling get_vegas_content()", plugin_id)
+            logger.debug("[%s] Native: calling get_vegas_content()", plugin_id)
 
             # Tell the plugin how much width the ticker wants it to use, and
             # narrow the canvas for the duration of the call. A plugin that
@@ -574,7 +707,7 @@ class PluginAdapter:
             # be explicit can read get_vegas_render_width().
             render_width = self.resolve_render_width(plugin, plugin_id)
             if render_width != self.display_width:
-                logger.info(
+                logger.debug(
                     "[%s] Native: requesting %dpx instead of %dpx",
                     plugin_id, render_width, self.display_width
                 )
@@ -602,19 +735,19 @@ class PluginAdapter:
                 plugin._vegas_render_width = None
 
             if result is None:
-                logger.info("[%s] Native: get_vegas_content() returned None", plugin_id)
+                logger.debug("[%s] Native: get_vegas_content() returned None", plugin_id)
                 return None
 
             # Normalize to list
             if isinstance(result, Image.Image):
                 images = [result]
-                logger.info(
+                logger.debug(
                     "[%s] Native: got single Image %dx%d",
                     plugin_id, result.width, result.height
                 )
             elif isinstance(result, (list, tuple)):
                 images = list(result)
-                logger.info(
+                logger.debug(
                     "[%s] Native: got %d items in list/tuple",
                     plugin_id, len(images)
                 )
@@ -635,14 +768,14 @@ class PluginAdapter:
                     )
                     continue
 
-                logger.info(
+                logger.debug(
                     "[%s] Native: item[%d] is %dx%d, mode=%s",
                     plugin_id, i, img.width, img.height, img.mode
                 )
 
                 # Ensure correct height
                 if img.height != self.display_height:
-                    logger.info(
+                    logger.debug(
                         "[%s] Native: resizing item[%d]: %dx%d -> %dx%d",
                         plugin_id, i, img.width, img.height,
                         img.width, self.display_height
@@ -660,13 +793,13 @@ class PluginAdapter:
 
             if valid_images:
                 total_width = sum(img.width for img in valid_images)
-                logger.info(
+                logger.debug(
                     "[%s] Native: SUCCESS - %d images, %dpx total width",
                     plugin_id, len(valid_images), total_width
                 )
                 return valid_images
 
-            logger.info("[%s] Native: no valid images after validation", plugin_id)
+            logger.debug("[%s] Native: no valid images after validation", plugin_id)
             return None
 
         except (AttributeError, TypeError, ValueError, OSError) as e:
@@ -700,20 +833,20 @@ class PluginAdapter:
                 logger.debug("[%s] No scroll_helper attribute", plugin_id)
                 return None
 
-            logger.info(
+            logger.debug(
                 "[%s] Found scroll_helper: %s",
                 plugin_id, type(scroll_helper).__name__
             )
 
             cached_image = getattr(scroll_helper, 'cached_image', None)
             if cached_image is None:
-                logger.info(
+                logger.debug(
                     "[%s] scroll_helper.cached_image is None, triggering content generation",
                     plugin_id
                 )
                 if offscreen_only:
                     # Generating it calls display(), which needs the canvas.
-                    logger.info(
+                    logger.debug(
                         "[%s] scroll_helper cache empty; deferring generation "
                         "to the render thread", plugin_id
                     )
@@ -726,13 +859,13 @@ class PluginAdapter:
                     return None
 
             if not isinstance(cached_image, Image.Image):
-                logger.info(
+                logger.debug(
                     "[%s] scroll_helper.cached_image is not an Image: %s",
                     plugin_id, type(cached_image).__name__
                 )
                 return None
 
-            logger.info(
+            logger.debug(
                 "[%s] scroll_helper.cached_image found: %dx%d, mode=%s",
                 plugin_id, cached_image.width, cached_image.height, cached_image.mode
             )
@@ -755,7 +888,7 @@ class PluginAdapter:
 
             # Ensure correct height
             if img.height != self.display_height:
-                logger.info(
+                logger.debug(
                     "[%s] Resizing scroll_helper content: %dx%d -> %dx%d",
                     plugin_id, img.width, img.height,
                     img.width, self.display_height
@@ -769,7 +902,7 @@ class PluginAdapter:
             if img.mode != 'RGB':
                 img = img.convert('RGB')
 
-            logger.info(
+            logger.debug(
                 "[%s] ScrollHelper content ready: %dx%d",
                 plugin_id, img.width, img.height
             )
@@ -869,7 +1002,7 @@ class PluginAdapter:
             with self._capture():
                 # Method 1: Try _create_scrolling_display (stocks pattern)
                 if hasattr(plugin, '_create_scrolling_display'):
-                    logger.info(
+                    logger.debug(
                         "[%s] Triggering via _create_scrolling_display()",
                         plugin_id
                     )
@@ -877,7 +1010,7 @@ class PluginAdapter:
                         plugin._create_scrolling_display()
                         cached_image = getattr(scroll_helper, 'cached_image', None)
                         if cached_image is not None and isinstance(cached_image, Image.Image):
-                            logger.info(
+                            logger.debug(
                                 "[%s] _create_scrolling_display() SUCCESS: %dx%d",
                                 plugin_id, cached_image.width, cached_image.height
                             )
@@ -889,7 +1022,7 @@ class PluginAdapter:
 
                 # Method 2: Try display(force_clear=True) which typically builds scroll content
                 if hasattr(plugin, 'display'):
-                    logger.info(
+                    logger.debug(
                         "[%s] Triggering via display(force_clear=True)",
                         plugin_id
                     )
@@ -898,12 +1031,12 @@ class PluginAdapter:
                         plugin.display(force_clear=True)
                         cached_image = getattr(scroll_helper, 'cached_image', None)
                         if cached_image is not None and isinstance(cached_image, Image.Image):
-                            logger.info(
+                            logger.debug(
                                 "[%s] display(force_clear=True) SUCCESS: %dx%d",
                                 plugin_id, cached_image.width, cached_image.height
                             )
                             return cached_image
-                        logger.info(
+                        logger.debug(
                             "[%s] display(force_clear=True) did not populate cached_image",
                             plugin_id
                         )
@@ -912,7 +1045,7 @@ class PluginAdapter:
                             "[%s] display(force_clear=True) failed", plugin_id
                         )
 
-            logger.info(
+            logger.debug(
                 "[%s] Could not trigger scroll content generation",
                 plugin_id
             )
@@ -944,15 +1077,15 @@ class PluginAdapter:
         try:
             # Save current display state
             original_image = self.display_manager.image.copy()
-            logger.info("[%s] Fallback: saved original display state", plugin_id)
+            logger.debug("[%s] Fallback: saved original display state", plugin_id)
 
             # Ensure plugin has fresh data before capturing
             has_update_data = hasattr(plugin, 'update_data')
-            logger.info("[%s] Fallback: has update_data=%s", plugin_id, has_update_data)
+            logger.debug("[%s] Fallback: has update_data=%s", plugin_id, has_update_data)
             if has_update_data:
                 try:
                     plugin.update_data()
-                    logger.info("[%s] Fallback: update_data() called", plugin_id)
+                    logger.debug("[%s] Fallback: update_data() called", plugin_id)
                 except (AttributeError, RuntimeError, OSError):
                     logger.exception("[%s] Fallback: update_data() failed", plugin_id)
 
@@ -964,41 +1097,41 @@ class PluginAdapter:
             # arrangement rather than one that has to be cropped afterwards.
             render_width = self.resolve_render_width(plugin, plugin_id)
             if render_width != self.display_width:
-                logger.info(
+                logger.debug(
                     "[%s] Fallback: rendering at %dpx instead of %dpx",
                     plugin_id, render_width, self.display_width
                 )
 
             with self._capture(), self._render_at(render_width):
                 self.display_manager.clear()
-                logger.info("[%s] Fallback: display cleared, calling display()", plugin_id)
+                logger.debug("[%s] Fallback: display cleared, calling display()", plugin_id)
 
                 # First try without force_clear (some plugins behave better this way)
                 try:
                     plugin.display()
-                    logger.info("[%s] Fallback: display() called successfully", plugin_id)
+                    logger.debug("[%s] Fallback: display() called successfully", plugin_id)
                 except TypeError:
                     # Plugin may require force_clear argument
-                    logger.info("[%s] Fallback: display() failed, trying with force_clear=True", plugin_id)
+                    logger.debug("[%s] Fallback: display() failed, trying with force_clear=True", plugin_id)
                     plugin.display(force_clear=True)
 
                 # Capture the result
                 captured = self.display_manager.image.copy()
 
-            logger.info(
+            logger.debug(
                 "[%s] Fallback: captured frame %dx%d, mode=%s",
                 plugin_id, captured.width, captured.height, captured.mode
             )
 
             # Check if captured image has content (not all black)
             is_blank, bright_ratio = self._is_blank_image(captured, return_ratio=True)
-            logger.info(
+            logger.debug(
                 "[%s] Fallback: brightness check - %.3f%% bright pixels (threshold=0.5%%)",
                 plugin_id, bright_ratio * 100
             )
 
             if is_blank:
-                logger.info(
+                logger.debug(
                     "[%s] Fallback: first capture blank, retrying with force_clear",
                     plugin_id
                 )
@@ -1009,7 +1142,7 @@ class PluginAdapter:
                     captured = self.display_manager.image.copy()
 
                 is_blank, bright_ratio = self._is_blank_image(captured, return_ratio=True)
-                logger.info(
+                logger.debug(
                     "[%s] Fallback: retry brightness - %.3f%% bright pixels",
                     plugin_id, bright_ratio * 100
                 )
@@ -1026,7 +1159,7 @@ class PluginAdapter:
             if captured.mode != 'RGB':
                 captured = captured.convert('RGB')
 
-            logger.info(
+            logger.debug(
                 "[%s] Fallback: SUCCESS - captured %dx%d",
                 plugin_id, captured.width, captured.height
             )

src/vegas_mode/stream_manager.py

@@ -406,6 +406,8 @@ class StreamManager:
         )
         logger.info("Ordered plugins: %s", ordered_plugins)
 
+        ordered_plugins = self._apply_priority_weights(ordered_plugins)
+
         # Atomically update shared state under lock to avoid races with prefetchers
         with self._buffer_lock:
             self._ordered_plugins = ordered_plugins
@@ -417,6 +419,143 @@ class StreamManager:
 
         logger.info("=" * 60)
 
+    def _plugin_weight(self, plugin_id: str) -> int:
+        """Slots per cycle for one plugin.
+
+        A plugin may answer for itself via get_vegas_priority_weight() -- the
+        only way favorite-team awareness can reach here, since the core can see
+        that a game is live but not whose. When it declines (returns None, the
+        default), live content earns ``live_weight`` and everything else 1.
+        """
+        plugin = None
+        try:
+            plugin = self.plugin_manager.plugins.get(plugin_id)
+        except (AttributeError, TypeError):
+            return 1
+        if plugin is None:
+            return 1
+
+        try:
+            if hasattr(plugin, 'get_vegas_priority_weight'):
+                declared = plugin.get_vegas_priority_weight()
+                if declared is not None:
+                    return max(1, min(10, int(declared)))
+        except Exception:
+            # Deliberately falls through to the core's own live check rather
+            # than demoting to 1. The plugin's weight calculation is broken,
+            # but has_live_priority() and has_live_content() are separate
+            # methods guarded separately below -- a plugin that genuinely has
+            # a live game should still get live_weight for it.
+            logger.exception("[%s] get_vegas_priority_weight() failed", plugin_id)
+
+        try:
+            if (hasattr(plugin, 'has_live_priority')
+                    and hasattr(plugin, 'has_live_content')
+                    and plugin.has_live_priority()
+                    and plugin.has_live_content()):
+                return self.config.live_weight
+        except Exception:
+            logger.exception("[%s] live-content check failed", plugin_id)
+        return 1
+
+    def _apply_priority_weights(self, ordered: List[str]) -> List[str]:
+        """Expand the rotation so weighted plugins take several turns per cycle.
+
+        Smooth Weighted Round-Robin, the same scheduler the sports plugins use
+        to rotate their own games: a plugin of weight N appears N times per
+        cycle, and the repeats are spaced through the cycle rather than
+        clumped, so a live score is never three-in-a-row followed by a long
+        silence.
+
+        Returns the input unchanged when nothing is weighted, which is both the
+        common case and the pre-existing behaviour.
+        """
+        if not ordered or not self.config.live_in_ticker:
+            return ordered
+
+        weights = {pid: self._plugin_weight(pid) for pid in ordered}
+        total = sum(weights.values())
+        if total <= len(ordered):
+            return ordered          # nothing boosted; plain round robin
+
+        current = {pid: 0 for pid in ordered}
+        schedule: List[str] = []
+        for _ in range(total):
+            for pid in ordered:
+                current[pid] += weights[pid]
+            picked = max(current, key=lambda p: current[p])
+            current[picked] -= total
+            schedule.append(picked)
+
+        schedule = self._unclump_seam(schedule)
+
+        boosted = {p: w for p, w in weights.items() if w > 1}
+        logger.info(
+            "Vegas rotation weighted: %d slots for %d plugins (boosted: %s)",
+            len(schedule), len(ordered), boosted)
+        return schedule
+
+    @staticmethod
+    def _unclump_seam(schedule: List[str]) -> List[str]:
+        """Stop the heaviest plugin sitting on both ends of the cycle.
+
+        Smooth Weighted Round-Robin spaces repeats well *within* a pass, but
+        it schedules the heaviest item first and often last too. The strip
+        loops, so those two are neighbours: the one place the marquee shows
+        the same plugin twice running is the seam between cycles.
+
+        Rotating the list cannot fix this. Rotation preserves the cyclic order
+        exactly, so it only moves where the seam is drawn, not the adjacency
+        itself. The trailing entry has to be swapped with one from the middle
+        whose neighbours differ from it, which breaks the pair without
+        creating another.
+
+        Left alone when no such position exists -- a rotation short enough or
+        lopsided enough to have none is one where the plugin is unavoidably
+        adjacent to itself anyway.
+        """
+        if len(schedule) < 3 or schedule[0] != schedule[-1]:
+            return schedule
+
+        repeated = schedule[-1]
+        size = len(schedule)
+
+        def cyclic_doubles(seq) -> int:
+            return sum(1 for i in range(size) if seq[i] == seq[(i + 1) % size])
+
+        def clearance(seq, value) -> int:
+            """Smallest cyclic gap between appearances of `value`."""
+            at = [i for i, v in enumerate(seq) if v == value]
+            if len(at) < 2:
+                return size
+            return min(min((b - a) % size, (a - b) % size)
+                       for i, a in enumerate(at) for b in at[i + 1:])
+
+        # Try each swap and judge the result, rather than reasoning about which
+        # neighbours the two moved elements will end up with. That reasoning is
+        # where the first version went wrong: it guarded the slot `repeated`
+        # moves into but not the one the displaced element lands in, so
+        # ['a','b','c','d','x','y','x','a'] came back ending ['x','x'] -- the
+        # seam duplicate traded for a fresh one.
+        best = None
+        best_clearance = -1
+        for j in range(1, size - 1):
+            candidate = list(schedule)
+            candidate[j], candidate[-1] = candidate[-1], candidate[j]
+            if cyclic_doubles(candidate):
+                continue
+            # Among the repairs that work, prefer the one that leaves the
+            # boosted plugin most evenly spread; taking the first that merely
+            # fits moved a repeat from a gap of 7 into a gap of 2.
+            spread = clearance(candidate, repeated)
+            if spread > best_clearance:
+                best, best_clearance = candidate, spread
+
+        # None exists when the value is unavoidably adjacent to itself -- a
+        # plugin holding most of the slots has to be. Schedule it as it is
+        # rather than refuse.
+        return best if best is not None else schedule
+
     def _prefetch_content(self, count: int = 1) -> None:
         """
         Prefetch content for upcoming plugins.

src/web_interface/api_helpers.py

@@ -30,16 +30,14 @@ def success_response(
     """
     response_data = create_success_response(data, message, metadata)
 
-    # Add request metadata if available
+    # Timing is merged into whatever the caller passed, without inventing a
-    if metadata is None:
+    # metadata block for responses that have neither.
-        metadata = {}
+    enriched = dict(metadata) if metadata is not None else {}
-    
-    # Add timing if request start time is available
     if hasattr(request, 'start_time'):
-        metadata['response_time_ms'] = int((time.time() - request.start_time) * 1000)
+        enriched['response_time_ms'] = int((time.time() - request.start_time) * 1000)
 
-    if metadata:
+    if metadata is not None or enriched:
-        response_data['metadata'] = metadata
+        response_data['metadata'] = enriched
 
     return jsonify(response_data)
 

src/web_interface/error_handler.py

@@ -4,6 +4,7 @@ Centralized error handling for web interface.
 Provides helpers for consistent error responses across API endpoints.
 """
 
+import re
 from typing import Any, Optional
 from flask import jsonify
 
@@ -16,6 +17,97 @@ from src.logging_config import get_logger
 logger = get_logger(__name__)
 
 
+# Credentials that turn up inside exception text. A requests error quotes the
+# URL it failed on, and plugins that authenticate by query string put their key
+# there, so echoing an exception verbatim can hand out an API key. Redact the
+# value, keep the parameter name -- knowing *which* credential was involved is
+# part of the diagnosis.
+_REDACT_CREDENTIAL = re.compile(
+    r'((?:api[_-]?key|access[_-]?token|auth|apikey|key|passwd|password|pwd|'
+    r'secret|sig|signature|token)["\']?\s*[=:]\s*["\']?)([^\s&"\'<>,}]+)',
+    re.IGNORECASE,
+)
+
+# `Authorization: <scheme> <credential>`. The scheme name is kept because it
+# says which kind of credential failed; the credential goes. Any scheme
+# matches, not a fixed list: ApiKey, Negotiate, NTLM, AWS4-HMAC-SHA256 and
+# whatever a plugin's API invents next are all credentials, and a list would
+# silently leak the ones nobody thought of. Not covered by the generic pattern
+# above, whose value part stops at whitespace and so would keep the credential
+# once a space follows the scheme.
+_REDACT_AUTH_HEADER = re.compile(
+    r'((?:proxy-)?authorization["\']?\s*[=:]\s*["\']?\s*'
+    r'(?:[A-Za-z][\w.+-]*[ \t]+)?)'          # optional scheme name, kept
+    r'([^\s,"\'<>}]+)',                       # the credential, redacted
+    re.IGNORECASE,
+)
+
+# Credentials embedded in a URL: https://user:password@host. requests quotes
+# the full URL in its exceptions, so this is a realistic leak. The username is
+# kept -- it identifies which account failed without being the secret.
+_REDACT_URL_USERINFO = re.compile(r'([a-z][a-z0-9+.-]*://[^/\s:@]+:)([^/\s@]+)(@)',
+                                  re.IGNORECASE)
+
+# Long enough for an errno string with a path, short enough not to dump a
+# parser's worth of context into a JSON field.
+_MAX_DETAIL_LENGTH = 400
+
+
+def describe_exception(exc: BaseException,
+                       max_length: int = _MAX_DETAIL_LENGTH) -> str:
+    """
+    One-line, safe-to-return description of an exception.
+
+    The generic "an error occurred; see logs for details" tells a user nothing
+    and, when the failure is bad enough, the logs are unreachable too: a device
+    whose storage was failing returned that message from every endpoint
+    *including* the log viewer, because journalctl could not be executed. The
+    underlying `[Errno 5] Input/output error` named the fault immediately.
+
+    Returns "TypeName: message", credentials redacted and length capped. The
+    type alone is worth carrying -- a bare PermissionError says more than any
+    generic sentence.
+
+    Args:
+        exc: The exception to describe
+        max_length: Truncate beyond this many characters
+
+    Returns:
+        A single-line description, never empty
+    """
+    message = str(exc).strip()
+    text = f"{type(exc).__name__}: {message}" if message else type(exc).__name__
+    return redact_text(text, max_length)
+
+
+def redact_text(text: str, max_length: int = _MAX_DETAIL_LENGTH) -> str:
+    """Make arbitrary text safe to hand back over HTTP.
+
+    Split out of describe_exception because exceptions are not the only thing
+    worth returning: a subprocess's stderr, or a message a helper script
+    printed, is just as useful to a user and just as capable of carrying a
+    token or a password in it.
+
+    Args:
+        text: The text to redact
+        max_length: Truncate beyond this many characters
+
+    Returns:
+        A single line, credentials replaced, length capped.
+    """
+    text = text or ''
+    # Order matters: the URL and header forms are more specific than the
+    # generic key=value pattern, which would otherwise chew the scheme.
+    text = _REDACT_URL_USERINFO.sub(r'\1<redacted>\3', text)
+    text = _REDACT_AUTH_HEADER.sub(r'\1<redacted>', text)
+    text = _REDACT_CREDENTIAL.sub(r'\1<redacted>', text)
+    # Collapse newlines/tabs so the detail stays one line in a JSON field.
+    text = ' '.join(text.split())
+    if len(text) > max_length:
+        text = text[:max_length - 1].rstrip() + '…'
+    return text
+
+
 def create_error_response(
     error_code: ErrorCode,
     message: str,
@@ -69,13 +161,16 @@ def create_success_response(
         "status": "success"
     }
     
+    # All three use `is not None` rather than truthiness: "" and {} are
+    # values a caller chose to send, and dropping them silently would make
+    # the response shape depend on the data.
     if data is not None:
         response["data"] = data
 
-    if message:
+    if message is not None:
         response["message"] = message
 
-    if metadata:
+    if metadata is not None:
         response["metadata"] = metadata
 
     return response

src/web_interface/errors.py

@@ -89,7 +89,11 @@ class WebInterfaceError:
         self.category = category or self._infer_category(error_code)
         self.details = details
         self.context = context or {}
-        self.suggested_fixes = suggested_fixes or self._get_default_suggestions(error_code)
+        # `is None`, not truthiness: an explicit [] means "this caller has
+        # no suggestions to offer", which the default list would override.
+        self.suggested_fixes = (
+            suggested_fixes if suggested_fixes is not None
+            else self._get_default_suggestions(error_code))
         self.original_error = original_error
     
     def _infer_category(self, error_code: ErrorCode) -> ErrorCategory:

src/web_interface/secret_helpers.py

@@ -5,7 +5,7 @@ Provides functions for identifying, masking, separating, and filtering
 secret fields in plugin configurations based on JSON Schema x-secret markers.
 """
 
-from typing import Any, Dict, Set, Tuple
+from typing import Any, Dict, Optional, Set, Tuple
 
 
 def find_secret_fields(properties: Dict[str, Any], prefix: str = '') -> Set[str]:
@@ -143,6 +143,12 @@ def mask_secret_fields(config: Dict[str, Any], schema_properties: Dict[str, Any]
     return result
 
 
+#: What a masked secret looks like on the wire. Named because the write path
+#: has to recognise it coming back: a client that renders the mask and posts
+#: it unchanged must not store the mask as if it were the secret.
+SECRET_MASK = '\u2022' * 8
+
+
 def mask_all_secret_values(config: Dict[str, Any]) -> Dict[str, Any]:
     """Blanket-mask every non-empty value in a secrets config dict.
 
@@ -156,15 +162,25 @@ def mask_all_secret_values(config: Dict[str, Any]) -> Dict[str, Any]:
     Returns:
         A copy with all real values replaced by ``'••••••••'``.
     """
-    masked: Dict[str, Any] = {}
+    return {k: _mask_value(v) for k, v in config.items()}
-    for k, v in config.items():
+
-        if isinstance(v, dict):
+
-            masked[k] = mask_all_secret_values(v)
+def _mask_value(value: Any) -> Any:
-        elif v not in (None, '') and not (isinstance(v, str) and v.startswith('YOUR_')):
+    """Mask one value, recursing through dicts and lists.
-            masked[k] = '••••••••'
+
-        else:
+    A list used to be masked as though it were a scalar, so
-            masked[k] = v
+    ``accounts: [{"name": "a", "token": "..."}]`` came back as a single
-    return masked
+    ``'••••••••'``. Nothing leaked, but the caller could no longer see how
+    many entries there were or any of their non-secret fields, and the raw
+    editor was shown a string where the file holds an array.
+    """
+    if isinstance(value, dict):
+        return {k: _mask_value(v) for k, v in value.items()}
+    if isinstance(value, list):
+        return [_mask_value(item) for item in value]
+    if value in (None, '') or (isinstance(value, str) and value.startswith('YOUR_')):
+        return value
+    return SECRET_MASK
 
 
 def remove_empty_secrets(secrets: Dict[str, Any]) -> Dict[str, Any]:
@@ -186,6 +202,133 @@ def remove_empty_secrets(secrets: Dict[str, Any]) -> Dict[str, Any]:
             nested = remove_empty_secrets(v)
             if nested:
                 result[k] = nested
+        elif isinstance(v, list):
+            # Lists used to fall through to the scalar branch below and be
+            # kept verbatim, blanks and all. Because lists merge by
+            # *replacement*, saving any unrelated setting then wrote
+            # [{"token": ""}, ...] straight over the stored list and
+            # destroyed every credential in it.
+            pruned = _prune_secret_list(v)
+            if pruned is not None:
+                result[k] = pruned
         elif v is not None and not (isinstance(v, str) and v.strip() == ''):
             result[k] = v
     return result
+
+
+def _prune_secret_list(items: list) -> Optional[list]:
+    """Strip blanks from inside a list of secrets, preserving every index.
+
+    The rest of the system treats a secrets list as *parallel* to the regular
+    one -- ``sec[i]`` holds the secret fields of item ``i``, and ``{}`` means
+    "item i has none" (see ConfigManager._strip_secrets_recursive). So an
+    emptied dict item stays ``{}``: putting ``None`` there makes that list stop
+    looking parallel, and the stripper then drops the whole key from the main
+    config, taking the non-secret fields with it.
+
+    A blank *scalar* becomes ``None``, meaning "no update at this index" --
+    :func:`merge_secrets` substitutes whatever is stored there. Returns
+    ``None`` when nothing in the list carries a real value, so the caller drops
+    the key and leaves the stored list untouched.
+    """
+    pruned: list = []
+    has_real_value = False
+    for item in items:
+        if isinstance(item, dict):
+            kept = remove_empty_secrets(item)
+            pruned.append(kept)
+            has_real_value = has_real_value or bool(kept)
+        elif isinstance(item, list):
+            sub = _prune_secret_list(item)
+            pruned.append(sub if sub is not None else [])
+            has_real_value = has_real_value or sub is not None
+        elif item is not None and not (isinstance(item, str) and item.strip() == ''):
+            pruned.append(item)
+            has_real_value = True
+        else:
+            pruned.append(None)
+    return pruned if has_real_value else None
+
+
+def merge_secrets(stored: Any, incoming: Any) -> Any:
+    """Merge submitted secrets over stored ones, element-wise inside lists.
+
+    ``deep_merge`` replaces a list wholesale. For secrets that is destructive:
+    an incoming list that carries a real value for one entry and ``None`` for
+    the rest would drop the stored credentials of every other entry. Here a
+    list merges by index, and ``None`` means "keep what is stored".
+
+    Entries are matched by *position*, which is what the config form gives us
+    -- there is no schema-declared identity to key on, and it is the same
+    contract ConfigManager._strip_secrets_recursive already relies on. The
+    incoming list's length wins, so deleting an item deletes its secrets;
+    an item the client left blank keeps whatever is stored at that index.
+    """
+    if isinstance(stored, dict) and isinstance(incoming, dict):
+        merged = dict(stored)
+        for key, value in incoming.items():
+            merged[key] = (merge_secrets(stored[key], value)
+                           if key in stored else value)
+        return merged
+    if isinstance(stored, list) and isinstance(incoming, list):
+        # The incoming list sets the length -- the regular config's list is
+        # authoritative about how many items exist, and this one runs parallel
+        # to it. Removing an entry must therefore remove its secrets too.
+        merged_list = []
+        for index, item in enumerate(incoming):
+            stored_item = stored[index] if index < len(stored) else None
+            merged_list.append(stored_item if item is None
+                               else merge_secrets(stored_item, item))
+        return merged_list
+    if incoming is None:
+        return stored
+    return incoming
+
+
+def strip_masked_values(secrets: Dict[str, Any]) -> Dict[str, Any]:
+    """Remove values a client echoed back rather than changed.
+
+    The counterpart to :func:`mask_all_secret_values`. A client that GETs the
+    masked secrets, edits one field and POSTs the whole object back is sending
+    ``SECRET_MASK`` for every field it did not touch. Storing those would
+    replace each untouched credential with eight bullet characters.
+
+    Drops the mask and, like :func:`remove_empty_secrets`, blank values -- so
+    the caller can merge the result onto what is already stored and have
+    "unchanged" mean unchanged. Empty nested dicts are pruned.
+    """
+    result: Dict[str, Any] = {}
+    for k, v in secrets.items():
+        if isinstance(v, dict):
+            nested = strip_masked_values(v)
+            if nested:
+                result[k] = nested
+        elif isinstance(v, list):
+            # A list is merged by replacement, not element by element -- there
+            # is no identity to match entries on -- so a list that still holds
+            # a mask cannot be merged safely: keeping it would store bullets,
+            # and keeping the submitted entries alone would drop whichever the
+            # client did not send back. Dropping the key leaves the stored
+            # list untouched, which is what an untouched list should do.
+            #
+            # The consequence, deliberately: editing one secret inside a list
+            # through this endpoint requires sending real values for all of
+            # them. Sending some masks leaves the whole list as it was.
+            if not _contains_mask(v):
+                result[k] = v
+        elif v is None:
+            continue
+        elif isinstance(v, str) and (v.strip() == '' or v == SECRET_MASK):
+            continue
+        else:
+            result[k] = v
+    return result
+
+
+def _contains_mask(value: Any) -> bool:
+    """True when a mask sentinel survives anywhere inside ``value``."""
+    if isinstance(value, dict):
+        return any(_contains_mask(v) for v in value.values())
+    if isinstance(value, list):
+        return any(_contains_mask(item) for item in value)
+    return value == SECRET_MASK

src/web_interface/validators.py

@@ -43,10 +43,15 @@ def validate_image_url(url: str) -> Tuple[bool, Optional[str]]:
     if any(handler in url_lower for handler in ['onerror=', 'onload=', 'onclick=']):
         return False, "Event handlers not allowed in URLs"
     
+    # Reject directory traversal anywhere, not only in relative paths:
+    # http://host/../secret is as much a traversal attempt as /../secret.
+    if '..' in url:
+        return False, "Invalid path: directory traversal not allowed"
+
     # Allow relative paths starting with /
     if url.startswith('/'):
-        # Validate it's a safe relative path (no directory traversal)
+        # // would be a protocol-relative URL, not a local path
-        if '..' in url or url.startswith('//'):
+        if url.startswith('//'):
             return False, "Invalid relative path"
         return True, None
     
@@ -104,10 +109,11 @@ def validate_file_upload(filename: str, max_size_mb: int = 10,
     if '..' in filename or '/' in filename or '\\' in filename:
         return False, "Filename contains invalid characters"
     
-    # Check extension if specified
+    # Check extension if specified. Both sides are lowercased: the caller's
+    # list is as likely to hold '.TTF' as the filename is.
     if allowed_extensions:
         file_ext = Path(filename).suffix.lower()
-        if file_ext not in allowed_extensions:
+        if file_ext not in [ext.lower() for ext in allowed_extensions]:
             return False, f"File extension must be one of: {', '.join(allowed_extensions)}"
     
     return True, None
@@ -147,7 +153,8 @@ def validate_numeric_range(value: float, min_val: Optional[float] = None,
     Returns:
         Tuple of (is_valid, error_message)
     """
-    if not isinstance(value, (int, float)):
+    # bool is an int subclass, so True would otherwise validate as 1.
+    if not isinstance(value, (int, float)) or isinstance(value, bool):
         return False, "Value must be a number"
     
     if min_val is not None and value < min_val:
@@ -183,7 +190,15 @@ def validate_string_length(text: str, min_length: Optional[int] = None,
 
 def sanitize_plugin_config(config: dict) -> dict:
     """
-    Sanitize plugin configuration input to prevent injection.
+    Restrict a plugin config to safe key names and value types.
+
+    Drops keys that are not plain identifiers and values that are not
+    JSON-ish scalars, lists, or dicts, recursing into the latter two.
+
+    String values are returned **unescaped**: output escaping is the
+    template layer's job, and escaping here would store the escaped form
+    in config.json. Do not read this function as XSS protection for
+    rendered output.
 
     Args:
         config: Configuration dictionary

systemd/ledmatrix-wifi-monitor.service

@@ -10,8 +10,8 @@ WorkingDirectory=__PROJECT_ROOT_DIR__
 ExecStart=/usr/bin/python3 __PROJECT_ROOT_DIR__/scripts/utils/wifi_monitor_daemon.py --interval 30
 Restart=on-failure
 RestartSec=10
-StandardOutput=syslog
+StandardOutput=journal
-StandardError=syslog
+StandardError=journal
 SyslogIdentifier=ledmatrix-wifi-monitor
 
 [Install]

systemd/ledmatrix.service

@@ -8,9 +8,38 @@ Type=simple
 User=root
 WorkingDirectory=__PROJECT_ROOT_DIR__
 Environment=PYTHONDONTWRITEBYTECODE=1
+# glibc gives each allocating thread its own malloc arena, up to 8 x CPU count,
+# and an arena that has grown is never handed back to the OS. This process runs
+# 9 threads on a 3-core Pi, so the ceiling is 24 arenas -- and a rig measured at
+# 1030 MB resident held 23 large anonymous mappings on 64 MB-aligned addresses,
+# 920 MB of them, while the live data it was actually holding (widest scroll
+# strip seen: 35,746 x 64) accounts for roughly 15 MB. That gap is arena bloat,
+# not leaked objects: RSS was flat across repeated sampling, not climbing.
+#
+# Capping the arenas trades a little allocator concurrency for a large amount of
+# resident memory on a device that has neither to spare. 2 is the usual value;
+# raise it if frame times regress.
+Environment=MALLOC_ARENA_MAX=2
 ExecStart=/usr/bin/python3 __PROJECT_ROOT_DIR__/run.py
-Restart=on-failure
+# Restart=always, not on-failure: run.py exiting 0 (a clean shutdown path taken
+# for a reason that no longer applies, e.g. a config reload) would otherwise leave
+# the service stopped and the panel dark indefinitely, with systemd considering
+# that a successful outcome and never bringing it back.
+Restart=always
 RestartSec=10
+# Memory ceiling as a share of physical RAM, so one unit file suits a 512 MB
+# Pi Zero 2 W and an 8 GB Pi 5 alike. This is a backstop, not a tuning knob: it
+# turns "the board runs out of memory, stops being able to fork, and takes sshd
+# and the panel down together until someone pulls the plug" into "this one
+# service restarts".
+#
+# NOTE: Raspberry Pi firmware boots the kernel with cgroup_disable=memory, and
+# systemd accepts this setting and then silently ignores it. Verify with:
+#     grep memory /sys/fs/cgroup/cgroup.controllers
+# If that prints nothing, add "cgroup_enable=memory cgroup_memory=1" to
+# /boot/firmware/cmdline.txt (all on line 1) and reboot. first_time_install.sh
+# does this for you.
+MemoryMax=85%
 StandardOutput=journal
 StandardError=journal
 SyslogIdentifier=ledmatrix

test/_api_v3_test_helpers.py

@@ -0,0 +1,75 @@
+"""
+Shared scaffolding for api_v3 blueprint tests.
+
+Not a test module (the leading underscore keeps pytest from collecting
+it). It is the pytest-fixture equivalent of ``_make_client()`` in
+test_uninstall_and_reconcile_endpoint.py, which is unittest-style and
+requires ``self.addCleanup``.
+
+The api_v3 blueprint keeps its managers as attributes on a module-level
+singleton, not in Flask app state, so replacing them with mocks leaks
+into every later test that imports api_v3 unless the originals are put
+back. ``api_v3_client`` snapshots and restores them around each test.
+"""
+
+from unittest.mock import MagicMock
+
+import pytest
+from flask import Flask
+
+
+# Every manager attribute the blueprint reads. Anything missing here keeps
+# whatever a previously-run test left on the singleton.
+API_V3_MANAGER_ATTRS = (
+    'config_manager', 'plugin_manager', 'plugin_store_manager',
+    'plugin_state_manager', 'saved_repositories_manager', 'schema_manager',
+    'operation_queue', 'operation_history', 'cache_manager',
+)
+
+_SENTINEL = object()
+
+
+def build_app(blueprint):
+    app = Flask(__name__)
+    app.config['TESTING'] = True
+    app.config['SECRET_KEY'] = 'test'
+    app.register_blueprint(blueprint, url_prefix='/api/v3')
+    return app
+
+
+@pytest.fixture
+def api_v3_module():
+    """The api_v3 module with every manager replaced by a MagicMock.
+
+    Restores the original attributes afterwards. Tests point individual
+    managers at real objects (a ConfigManager over tmp_path, say) or set
+    them to None to exercise the not-initialized branches.
+    """
+    from web_interface.blueprints import api_v3 as module
+
+    originals = {
+        name: getattr(module.api_v3, name, _SENTINEL)
+        for name in API_V3_MANAGER_ATTRS
+    }
+    for name in API_V3_MANAGER_ATTRS:
+        setattr(module.api_v3, name, MagicMock())
+    # Default to the direct path; queue tests opt in explicitly.
+    module.api_v3.operation_queue = None
+
+    yield module
+
+    for name, original in originals.items():
+        if original is _SENTINEL:
+            if hasattr(module.api_v3, name):
+                try:
+                    delattr(module.api_v3, name)
+                except AttributeError:
+                    pass
+        else:
+            setattr(module.api_v3, name, original)
+
+
+@pytest.fixture
+def api_v3_client(api_v3_module):
+    """Flask test client wired to the mocked blueprint."""
+    return build_app(api_v3_module.api_v3).test_client()

test/test_api_helper.py

@@ -0,0 +1,275 @@
+"""
+Tests for src/common/api_helper.py (APIHelper).
+
+Covers rate limiting, cached GETs, ESPN URL/cache-key construction,
+session header defaults and per-call merging, the retry adapter, and the
+fixed clear_cache() behavior (real CacheManager surface: clear_cache /
+delete / list_cache_files, with safe no-ops elsewhere).
+
+No real network: helper.session.get/post are always replaced with mocks.
+"""
+
+import types
+from unittest.mock import MagicMock, Mock
+
+import pytest
+import requests
+from freezegun import freeze_time
+
+import src.common.api_helper as api_helper_module
+from src.common.api_helper import APIHelper
+
+
+def _make_response(payload):
+    response = MagicMock()
+    response.json.return_value = payload
+    response.raise_for_status.return_value = None
+    return response
+
+
+@pytest.fixture
+def cache():
+    cache = MagicMock()
+    cache.get.return_value = None
+    return cache
+
+
+@pytest.fixture
+def helper(cache):
+    helper = APIHelper(cache_manager=cache)
+    # Default min interval is 1.0s and would really sleep between requests.
+    helper.set_rate_limit(0)
+    return helper
+
+
+# ---------------------------------------------------------------------------
+# Rate limiting
+# ---------------------------------------------------------------------------
+
+class TestRateLimiting:
+    def test_sleeps_for_remaining_interval(self, helper, monkeypatch):
+        fake_time = MagicMock()
+        fake_time.time.side_effect = [102.0, 105.0]
+        monkeypatch.setattr(api_helper_module, 'time', fake_time)
+
+        helper.set_rate_limit(5)
+        helper._last_request_time = 100.0
+        helper._enforce_rate_limit()
+
+        # 2s elapsed of a 5s interval -> sleep the remaining 3s.
+        fake_time.sleep.assert_called_once()
+        assert fake_time.sleep.call_args[0][0] == pytest.approx(3.0)
+        assert helper._last_request_time == 105.0
+
+    def test_no_sleep_when_interval_elapsed(self, helper, monkeypatch):
+        fake_time = MagicMock()
+        fake_time.time.side_effect = [200.0, 201.0]
+        monkeypatch.setattr(api_helper_module, 'time', fake_time)
+
+        helper.set_rate_limit(5)
+        helper._last_request_time = 100.0
+        helper._enforce_rate_limit()
+
+        fake_time.sleep.assert_not_called()
+        assert helper._last_request_time == 201.0
+
+
+# ---------------------------------------------------------------------------
+# get()
+# ---------------------------------------------------------------------------
+
+class TestGet:
+    def test_cache_hit_skips_request_and_rate_limit(self, helper, cache):
+        cache.get.return_value = {'cached': True}
+        helper.session.get = Mock()
+        rate_spy = Mock()
+        helper._enforce_rate_limit = rate_spy
+
+        result = helper.get('https://example.com/api', cache_key='k')
+
+        assert result == {'cached': True}
+        helper.session.get.assert_not_called()
+        rate_spy.assert_not_called()
+
+    def test_cache_miss_fetches_and_caches_without_ttl(self, helper, cache):
+        cache.get.return_value = None
+        helper.session.get = Mock(return_value=_make_response({'a': 1}))
+
+        result = helper.get('https://example.com/api', cache_key='k',
+                            cache_ttl=999)
+
+        assert result == {'a': 1}
+        # Pin the ttl-dropped contract: CacheManager.set is called with
+        # (key, data) only — the cache_ttl argument is discarded.
+        cache.set.assert_called_once_with('k', {'a': 1})
+
+    def test_request_exception_returns_none_and_caches_nothing(
+            self, helper, cache):
+        helper.session.get = Mock(
+            side_effect=requests.exceptions.RequestException('boom'))
+
+        result = helper.get('https://example.com/api', cache_key='k')
+
+        assert result is None
+        cache.set.assert_not_called()
+
+    def test_timeout_zero_falls_back_to_default(self, helper):
+        # Quirk pin: `timeout or self.default_timeout` treats an explicit
+        # timeout=0 as falsy, so the default (30) is used instead.
+        helper.session.get = Mock(return_value=_make_response({}))
+
+        helper.get('https://example.com/api', timeout=0)
+
+        assert helper.session.get.call_args.kwargs['timeout'] == 30
+
+    def test_per_call_headers_merge_over_session_headers(self, helper):
+        helper.session.get = Mock(return_value=_make_response({}))
+
+        helper.get('https://example.com/api', headers={'X-Custom': 'yes'})
+
+        sent = helper.session.get.call_args.kwargs['headers']
+        # Merged, not replaced: session defaults survive alongside the
+        # per-call header.
+        assert sent['X-Custom'] == 'yes'
+        assert sent['User-Agent'] == (
+            'LEDMatrix/1.0 (+https://github.com/ChuckBuilds/LEDMatrix)')
+        assert sent['Accept'] == 'application/json'
+        # The session's own headers are not polluted by the per-call ones.
+        assert 'X-Custom' not in helper.session.headers
+
+
+# ---------------------------------------------------------------------------
+# ESPN helpers
+# ---------------------------------------------------------------------------
+
+class TestEspnHelpers:
+    @freeze_time('2026-08-07')
+    def test_fetch_espn_scoreboard_url_params_and_cache_key(self, helper):
+        helper.get = Mock(return_value={'ok': 1})
+
+        result = helper.fetch_espn_scoreboard('football', 'nfl')
+
+        assert result == {'ok': 1}
+        helper.get.assert_called_once_with(
+            'https://site.api.espn.com/apis/site/v2/sports/football/nfl/scoreboard',
+            params={'dates': '20260807', 'limit': 1000},
+            cache_key='espn_football_nfl_20260807',
+            cache_ttl=300,
+        )
+
+    def test_fetch_espn_scoreboard_explicit_date(self, helper):
+        helper.get = Mock(return_value=None)
+
+        helper.fetch_espn_scoreboard('basketball', 'nba', date='20250115')
+
+        kwargs = helper.get.call_args.kwargs
+        assert kwargs['params'] == {'dates': '20250115', 'limit': 1000}
+        assert kwargs['cache_key'] == 'espn_basketball_nba_20250115'
+
+    def test_fetch_espn_standings_url_and_cache_key(self, helper):
+        helper.get = Mock(return_value={'ok': 1})
+
+        helper.fetch_espn_standings('football', 'nfl')
+
+        helper.get.assert_called_once_with(
+            'https://site.api.espn.com/apis/site/v2/sports/football/nfl/standings',
+            cache_key='espn_standings_football_nfl',
+            cache_ttl=3600,
+        )
+
+    def test_fetch_espn_rankings_url_and_cache_key(self, helper):
+        helper.get = Mock(return_value={'ok': 1})
+
+        helper.fetch_espn_rankings('football', 'college-football')
+
+        helper.get.assert_called_once_with(
+            'https://site.api.espn.com/apis/site/v2/sports/football/college-football/rankings',
+            cache_key='espn_rankings_football_college-football',
+            cache_ttl=3600,
+        )
+
+
+# ---------------------------------------------------------------------------
+# Session setup
+# ---------------------------------------------------------------------------
+
+class TestSessionSetup:
+    def test_user_agent_exact(self, helper):
+        # Regression guard: ESPN began 403ing other user agents; this exact
+        # string must be sent on every request.
+        assert helper.session.headers['User-Agent'] == (
+            'LEDMatrix/1.0 (+https://github.com/ChuckBuilds/LEDMatrix)')
+
+    def test_retry_adapter_configuration(self):
+        helper = APIHelper(cache_manager=None, max_retries=7)
+
+        retries = helper.session.get_adapter('https://x').max_retries
+        assert retries.total == 7
+        assert {429, 500, 502, 503, 504} <= set(retries.status_forcelist)
+
+
+# ---------------------------------------------------------------------------
+# clear_cache (fixed behavior: real CacheManager surface)
+# ---------------------------------------------------------------------------
+
+class TestClearCache:
+    def test_no_pattern_uses_clear_cache_method(self):
+        manager = types.SimpleNamespace(clear_cache=Mock())
+        helper = APIHelper(cache_manager=manager)
+        helper.set_rate_limit(0)
+
+        helper.clear_cache()
+
+        manager.clear_cache.assert_called_once_with()
+
+    def test_no_pattern_falls_back_to_clear(self):
+        manager = types.SimpleNamespace(clear=Mock())
+        helper = APIHelper(cache_manager=manager)
+        helper.set_rate_limit(0)
+
+        helper.clear_cache()
+
+        manager.clear.assert_called_once_with()
+
+    def test_no_pattern_manager_without_any_clear_is_noop(self):
+        helper = APIHelper(cache_manager=object())
+        helper.set_rate_limit(0)
+
+        helper.clear_cache()  # must not raise
+
+    def test_pattern_deletes_only_matching_keys(self):
+        manager = types.SimpleNamespace(
+            list_cache_files=Mock(return_value=[
+                {'key': 'espn_nfl_x'},
+                {'key': 'other'},
+            ]),
+            delete=Mock(),
+        )
+        helper = APIHelper(cache_manager=manager)
+        helper.set_rate_limit(0)
+
+        helper.clear_cache(pattern='espn')
+
+        manager.delete.assert_called_once_with('espn_nfl_x')
+
+    def test_pattern_manager_without_list_cache_files_is_noop(self):
+        helper = APIHelper(cache_manager=object())
+        helper.set_rate_limit(0)
+
+        helper.clear_cache(pattern='espn')  # must not raise
+
+
+# ---------------------------------------------------------------------------
+# No cache manager
+# ---------------------------------------------------------------------------
+
+class TestNoCacheManager:
+    def test_all_cache_operations_safe_without_manager(self):
+        helper = APIHelper(cache_manager=None)
+        helper.set_rate_limit(0)
+
+        assert helper.get_cache('k') is None
+        assert helper._get_from_cache('k') is None
+        assert helper.set_cache('k', {'a': 1}) is None
+        assert helper.clear_cache() is None
+        assert helper.clear_cache(pattern='espn') is None

test/test_api_v3_calendar_credentials.py

@@ -0,0 +1,226 @@
+"""
+Endpoint tests for POST /plugins/calendar/upload-credentials.
+
+The endpoint takes an uploaded Google OAuth credentials file, writes it
+into the calendar plugin's directory as credentials.json at mode 0600, and
+copies any previous file aside first. It had no tests.
+
+Regression coverage for two fixed bugs:
+- The OAuth-shape check sat inside `except Exception: pass`, so a valid
+  JSON document that is not an object — a bare `42`, a list, a string —
+  raised TypeError on the membership test, was swallowed, and got saved
+  as credentials.json anyway.
+- Each overwrite created a timestamped backup and nothing ever removed
+  them, so every re-upload left another complete copy of the user's OAuth
+  client credentials in the plugin directory, indefinitely.
+"""
+
+import io
+import json
+import os
+import stat
+import sys
+import time
+from pathlib import Path
+from types import SimpleNamespace
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from test._api_v3_test_helpers import api_v3_client, api_v3_module  # noqa: F401,E402
+
+URL = "/api/v3/plugins/calendar/upload-credentials"
+
+VALID_CREDENTIALS = {
+    "installed": {
+        "client_id": "abc.apps.googleusercontent.com",
+        "client_secret": "shh",
+        "redirect_uris": ["http://localhost"],
+    }
+}
+
+
+@pytest.fixture
+def plugin_dir(tmp_path, api_v3_module):
+    directory = tmp_path / "plugins" / "calendar"
+    directory.mkdir(parents=True)
+    api_v3_module.api_v3.plugin_manager.get_plugin_directory.return_value = str(directory)
+    return directory
+
+
+def upload(client, content, filename="credentials.json"):
+    # bytes are sent verbatim (to exercise malformed input); anything else
+    # is serialized, so None becomes the JSON literal null rather than an
+    # empty body.
+    payload = content if isinstance(content, bytes) else json.dumps(content).encode()
+    return client.post(
+        URL,
+        data={"file": (io.BytesIO(payload), filename)},
+        content_type="multipart/form-data",
+    )
+
+
+def backups(plugin_dir):
+    return sorted(plugin_dir.glob("credentials.json.backup.*"))
+
+
+class TestRequestValidation:
+    def test_no_file_part_is_a_400(self, api_v3_client, plugin_dir):
+        response = api_v3_client.post(URL, data={}, content_type="multipart/form-data")
+        assert response.status_code == 400
+        assert "No file provided" in response.get_json()["message"]
+
+    def test_empty_filename_is_a_400(self, api_v3_client, plugin_dir):
+        response = upload(api_v3_client, VALID_CREDENTIALS, filename="")
+        assert response.status_code == 400
+
+    @pytest.mark.parametrize("filename", ["creds.txt", "creds.pem", "creds"])
+    def test_non_json_extension_is_a_400(self, api_v3_client, plugin_dir, filename):
+        response = upload(api_v3_client, VALID_CREDENTIALS, filename=filename)
+        assert response.status_code == 400
+        assert "JSON file" in response.get_json()["message"]
+
+    def test_uppercase_json_extension_accepted(self, api_v3_client, plugin_dir):
+        assert upload(api_v3_client, VALID_CREDENTIALS,
+                      filename="CREDENTIALS.JSON").status_code == 200
+
+    def test_oversized_file_is_a_400(self, api_v3_client, plugin_dir):
+        response = upload(api_v3_client, b"x" * (1024 * 1024 + 1))
+        assert response.status_code == 400
+        assert "1MB" in response.get_json()["message"]
+        assert not (plugin_dir / "credentials.json").exists()
+
+    def test_invalid_json_is_a_400(self, api_v3_client, plugin_dir):
+        response = upload(api_v3_client, b"{not json")
+        assert response.status_code == 400
+        assert "not valid JSON" in response.get_json()["message"]
+        assert not (plugin_dir / "credentials.json").exists()
+
+    def test_missing_plugin_directory_is_a_404(self, api_v3_client, api_v3_module, tmp_path):
+        api_v3_module.api_v3.plugin_manager.get_plugin_directory.return_value = str(
+            tmp_path / "not-installed")
+        assert upload(api_v3_client, VALID_CREDENTIALS).status_code == 404
+
+
+class TestOAuthShapeValidation:
+    def test_installed_key_accepted(self, api_v3_client, plugin_dir):
+        assert upload(api_v3_client, VALID_CREDENTIALS).status_code == 200
+
+    def test_web_key_accepted(self, api_v3_client, plugin_dir):
+        assert upload(api_v3_client, {"web": {"client_id": "x"}}).status_code == 200
+
+    def test_object_without_oauth_keys_is_a_400(self, api_v3_client, plugin_dir):
+        response = upload(api_v3_client, {"something": "else"})
+        assert response.status_code == 400
+        assert "valid Google OAuth" in response.get_json()["message"]
+        assert not (plugin_dir / "credentials.json").exists()
+
+    @pytest.mark.parametrize("content", [42, "a string", [1, 2, 3], True, None])
+    def test_valid_json_that_is_not_an_object_is_rejected(
+            self, api_v3_client, plugin_dir, content):
+        # Regression: `'installed' not in 42` raises TypeError, which the
+        # bare `except Exception: pass` swallowed — the file was then saved
+        # as credentials.json despite being unusable as credentials.
+        response = upload(api_v3_client, content)
+        assert response.status_code == 400
+        assert "valid Google OAuth" in response.get_json()["message"]
+        assert not (plugin_dir / "credentials.json").exists()
+
+
+class TestSaving:
+    def test_file_written_with_contents_intact(self, api_v3_client, plugin_dir):
+        response = upload(api_v3_client, VALID_CREDENTIALS)
+        assert response.status_code == 200
+        saved = json.loads((plugin_dir / "credentials.json").read_text())
+        assert saved == VALID_CREDENTIALS
+
+    def test_response_reports_the_path(self, api_v3_client, plugin_dir):
+        body = upload(api_v3_client, VALID_CREDENTIALS).get_json()
+        assert body["path"].endswith("credentials.json")
+
+    def test_permissions_are_owner_only(self, api_v3_client, plugin_dir):
+        upload(api_v3_client, VALID_CREDENTIALS)
+        mode = stat.S_IMODE((plugin_dir / "credentials.json").stat().st_mode)
+        assert mode == 0o600
+
+    def test_first_upload_creates_no_backup(self, api_v3_client, plugin_dir):
+        upload(api_v3_client, VALID_CREDENTIALS)
+        assert backups(plugin_dir) == []
+
+    def test_overwrite_backs_up_the_previous_file(self, api_v3_client, plugin_dir):
+        (plugin_dir / "credentials.json").write_text(json.dumps({"installed": {"old": 1}}))
+        upload(api_v3_client, VALID_CREDENTIALS)
+        assert len(backups(plugin_dir)) == 1
+        assert json.loads(backups(plugin_dir)[0].read_text()) == {"installed": {"old": 1}}
+        assert json.loads((plugin_dir / "credentials.json").read_text()) == VALID_CREDENTIALS
+
+
+class TestBackupPruning:
+    def _seed(self, plugin_dir, count):
+        """Create `count` backups with distinct, increasing mtimes."""
+        now = int(time.time())
+        for i in range(count):
+            path = plugin_dir / f"credentials.json.backup.{now - (count - i) * 10}"
+            path.write_text(json.dumps({"installed": {"gen": i}}))
+            os.utime(path, (now - (count - i) * 10, now - (count - i) * 10))
+
+    def test_old_backups_are_pruned(self, api_v3_client, plugin_dir):
+        # Regression: nothing ever removed these, so a plugin directory
+        # accumulated one full copy of the user's OAuth credentials per
+        # re-upload, forever.
+        (plugin_dir / "credentials.json").write_text(json.dumps({"installed": {"cur": 1}}))
+        self._seed(plugin_dir, 7)
+        assert len(backups(plugin_dir)) == 7
+
+        upload(api_v3_client, VALID_CREDENTIALS)
+        assert len(backups(plugin_dir)) == 5
+
+    def test_the_newest_backups_are_the_ones_kept(self, api_v3_client, plugin_dir):
+        (plugin_dir / "credentials.json").write_text(json.dumps({"installed": {"cur": 1}}))
+        self._seed(plugin_dir, 7)
+
+        upload(api_v3_client, VALID_CREDENTIALS)
+        remaining = backups(plugin_dir)
+        # The just-created backup (of "cur") plus the four newest seeds.
+        contents = [json.loads(p.read_text()) for p in remaining]
+        assert {"installed": {"cur": 1}} in contents
+        assert {"installed": {"gen": 0}} not in contents  # oldest seed gone
+
+    def test_under_the_limit_nothing_is_removed(self, api_v3_client, plugin_dir):
+        (plugin_dir / "credentials.json").write_text(json.dumps({"installed": {"cur": 1}}))
+        self._seed(plugin_dir, 2)
+        upload(api_v3_client, VALID_CREDENTIALS)
+        assert len(backups(plugin_dir)) == 3  # 2 seeded + 1 new
+
+    def test_repeated_uploads_stay_bounded(
+            self, api_v3_client, plugin_dir, api_v3_module, monkeypatch):
+        # The backup filename carries int(time.time()), so uploads inside
+        # the same second all write the same name and overwrite each other.
+        # Advance a fake clock a second per round — otherwise this never
+        # reaches six backups and the bound holds for the wrong reason.
+        clock = {"now": int(time.time())}
+        monkeypatch.setattr(
+            api_v3_module, "time", SimpleNamespace(time=lambda: clock["now"]))
+        for i in range(10):
+            clock["now"] += 1
+            upload(api_v3_client, {"installed": {"round": i}})
+            os.utime(plugin_dir / "credentials.json",
+                     (clock["now"], clock["now"]))
+        remaining = backups(plugin_dir)
+        assert len(remaining) == 5
+        # And they are the five most recent rounds, not an arbitrary five.
+        kept = sorted(int(p.name.rsplit(".", 1)[1]) for p in remaining)
+        assert kept == [clock["now"] - 4 + i for i in range(5)]
+
+    def test_unremovable_backup_does_not_fail_the_upload(
+            self, api_v3_client, plugin_dir, monkeypatch):
+        (plugin_dir / "credentials.json").write_text(json.dumps({"installed": {"cur": 1}}))
+        self._seed(plugin_dir, 7)
+
+        def refuse(self):
+            raise OSError("read-only filesystem")
+        monkeypatch.setattr(Path, "unlink", refuse)
+
+        # Pruning is housekeeping; failing it must not lose the upload.
+        assert upload(api_v3_client, VALID_CREDENTIALS).status_code == 200

test/test_api_v3_music_auth_endpoints.py

@@ -0,0 +1,302 @@
+"""
+Endpoint tests for /plugins/authenticate/spotify and .../ytm.
+
+The Spotify step-2 handler writes a Python wrapper script to a temp file
+with the user's redirect URL embedded in it, then runs that file through
+subprocess. That is the most dangerous shape in the blueprint and had no
+tests: the URL is user input reaching generated source code.
+
+The two endpoints are NOT symmetrical, despite the matching names. Only
+Spotify has a two-step flow, a wrapper script, and a redirect_url; YTM
+just runs its script directly.
+
+Regression coverage for one fixed bug: the wrapper file was unlinked in
+the success/failure branch and again in the TimeoutExpired handler, so
+any other failure from subprocess.run — the interpreter missing, a fork
+failure, an interrupted call — left a temp file containing the user's
+redirect URL behind.
+"""
+
+import ast
+import json
+import os
+import subprocess
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from test._api_v3_test_helpers import api_v3_client, api_v3_module  # noqa: F401,E402
+
+
+@pytest.fixture
+def plugin_dir(tmp_path, api_v3_module):
+    """A plugin directory containing both auth scripts."""
+    directory = tmp_path / "plugins" / "ledmatrix-music"
+    directory.mkdir(parents=True)
+    (directory / "authenticate_spotify.py").write_text("print('spotify')\n")
+    (directory / "authenticate_ytm.py").write_text("print('ytm')\n")
+    api_v3_module.api_v3.plugin_manager.get_plugin_directory.return_value = str(directory)
+    return directory
+
+
+def completed(returncode=0, stdout="ok", stderr=""):
+    return subprocess.CompletedProcess(
+        args=["python3"], returncode=returncode, stdout=stdout, stderr=stderr)
+
+
+class TestSpotifyPreconditions:
+    URL = "/api/v3/plugins/authenticate/spotify"
+
+    def test_missing_plugin_directory_is_404(self, api_v3_client, api_v3_module, tmp_path):
+        api_v3_module.api_v3.plugin_manager.get_plugin_directory.return_value = str(
+            tmp_path / "not-installed")
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 404
+        assert response.get_json()["message"] == "Plugin not found"
+
+    def test_none_plugin_directory_is_404(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_manager.get_plugin_directory.return_value = None
+        assert api_v3_client.post(self.URL, json={}).status_code == 404
+
+    def test_missing_auth_script_is_404(self, api_v3_client, plugin_dir):
+        (plugin_dir / "authenticate_spotify.py").unlink()
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 404
+        assert "script not found" in response.get_json()["message"]
+
+
+class TestSpotifyStepTwo:
+    """redirect_url present — the wrapper-script path."""
+
+    URL = "/api/v3/plugins/authenticate/spotify"
+
+    def test_success(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed(0, "done")):
+            response = api_v3_client.post(self.URL, json={"redirect_url": "http://cb/?code=x"})
+        assert response.status_code == 200
+        body = response.get_json()
+        assert body["status"] == "success"
+        assert body["output"] == "done"
+
+    def test_script_failure_is_a_400_with_combined_output(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed(1, "out", "err")):
+            response = api_v3_client.post(self.URL, json={"redirect_url": "http://cb/?code=x"})
+        assert response.status_code == 400
+        assert response.get_json()["output"] == "outerr"
+
+    def test_timeout_is_a_408(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run",
+                          side_effect=subprocess.TimeoutExpired("python3", 120)):
+            response = api_v3_client.post(self.URL, json={"redirect_url": "http://cb/?code=x"})
+        assert response.status_code == 408
+        assert "timed out" in response.get_json()["message"]
+
+    def test_runs_a_list_argv_never_a_shell(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed()) as run:
+            api_v3_client.post(self.URL, json={"redirect_url": "http://cb/?code=x"})
+        args, kwargs = run.call_args
+        assert isinstance(args[0], list)
+        assert args[0][0] == "python3"
+        assert kwargs.get("shell") in (None, False)
+
+    def test_timeout_is_bounded(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed()) as run:
+            api_v3_client.post(self.URL, json={"redirect_url": "http://cb/?code=x"})
+        assert run.call_args.kwargs["timeout"] == 120
+
+
+class TestSpotifyWrapperCleanup:
+    URL = "/api/v3/plugins/authenticate/spotify"
+
+    def _wrapper_paths_after(self, api_v3_client, run_mock):
+        """Run the endpoint and return the wrapper path subprocess saw."""
+        seen = {}
+
+        def capture(args, **kwargs):
+            seen["path"] = args[1]
+            return run_mock(args, **kwargs)
+
+        with patch.object(subprocess, "run", side_effect=capture):
+            api_v3_client.post(self.URL, json={"redirect_url": "http://cb/?code=x"})
+        return seen["path"]
+
+    def test_removed_after_success(self, api_v3_client, plugin_dir):
+        path = self._wrapper_paths_after(api_v3_client, lambda *a, **kw: completed())
+        assert not os.path.exists(path)
+
+    def test_removed_after_script_failure(self, api_v3_client, plugin_dir):
+        path = self._wrapper_paths_after(
+            api_v3_client, lambda *a, **kw: completed(1, "out", "err"))
+        assert not os.path.exists(path)
+
+    def test_removed_after_timeout(self, api_v3_client, plugin_dir):
+        def raise_timeout(*a, **kw):
+            raise subprocess.TimeoutExpired("python3", 120)
+        path = self._wrapper_paths_after(api_v3_client, raise_timeout)
+        assert not os.path.exists(path)
+
+    def test_removed_when_subprocess_cannot_start(self, api_v3_client, plugin_dir):
+        # Regression: cleanup lived in the success/failure branch and in the
+        # TimeoutExpired handler only. An OSError from subprocess.run itself
+        # — no interpreter, fork failure — skipped both and left the wrapper,
+        # which contains the user's redirect URL, on disk.
+        def raise_oserror(*a, **kw):
+            raise OSError("[Errno 12] Cannot allocate memory")
+        path = self._wrapper_paths_after(api_v3_client, raise_oserror)
+        assert not os.path.exists(path)
+
+
+class TestSpotifyRedirectUrlIsNotInjectable:
+    """The wrapper embeds redirect_url into generated Python source."""
+
+    URL = "/api/v3/plugins/authenticate/spotify"
+
+    ADVERSARIAL = [
+        '''http://cb/?code=x"''',
+        """http://cb/?code=x'""",
+        'http://cb/?code=x\\',
+        'http://cb/?code=x\nimport os; os.system("id")',
+        'http://cb/?code=x"""\nimport os\n"""',
+        "http://cb/?code=x'''",
+        'http://cb/?code=x\\"\\n',
+        '"; import os; os.system("id"); "',
+    ]
+
+    def _wrapper_source(self, api_v3_client, redirect_url):
+        captured = {}
+
+        def capture(args, **kwargs):
+            captured["source"] = Path(args[1]).read_text()
+            return completed()
+
+        with patch.object(subprocess, "run", side_effect=capture):
+            api_v3_client.post(self.URL, json={"redirect_url": redirect_url})
+        return captured["source"]
+
+    @pytest.mark.parametrize("redirect_url", ADVERSARIAL)
+    def test_wrapper_is_still_valid_python(self, api_v3_client, plugin_dir, redirect_url):
+        # If escaping failed, the generated file would not parse at all.
+        source = self._wrapper_source(api_v3_client, redirect_url)
+        ast.parse(source)
+
+    @pytest.mark.parametrize("redirect_url", ADVERSARIAL)
+    def test_url_survives_as_one_string_literal(
+            self, api_v3_client, plugin_dir, redirect_url):
+        # Stronger than "it parses": the URL must still be a single string
+        # assigned to redirect_url, not code that escaped into statements.
+        source = self._wrapper_source(api_v3_client, redirect_url)
+        tree = ast.parse(source)
+        assigned = [
+            node.value.value for node in ast.walk(tree)
+            if isinstance(node, ast.Assign)
+            and isinstance(node.value, ast.Constant)
+            and any(getattr(t, "id", None) == "redirect_url" for t in node.targets)
+        ]
+        assert assigned == [redirect_url.strip()]
+
+    def test_injected_call_does_not_become_a_statement(self, api_v3_client, plugin_dir):
+        source = self._wrapper_source(
+            api_v3_client, 'http://cb/\nimport os; os.system("id")')
+        tree = ast.parse(source)
+        imported = {
+            alias.name for node in ast.walk(tree)
+            if isinstance(node, ast.Import) for alias in node.names
+        }
+        # The wrapper legitimately imports sys, subprocess and os; what it
+        # must not gain is a *call* smuggled in through the URL.
+        calls = [
+            node for node in ast.walk(tree)
+            if isinstance(node, ast.Call)
+            and isinstance(node.func, ast.Attribute)
+            and node.func.attr == "system"
+        ]
+        assert calls == []
+
+
+class TestSpotifyStepOne:
+    """No redirect_url — the OAuth-URL path, which imports the script."""
+
+    URL = "/api/v3/plugins/authenticate/spotify"
+
+    def test_script_without_credentials_helper_is_an_error(
+            self, api_v3_client, plugin_dir):
+        # The stub script defines neither get_auth_url nor
+        # load_spotify_credentials, so no URL can be produced.
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code in (400, 500)
+        assert response.get_json()["status"] == "error"
+
+    def test_unusable_credentials_do_not_leak_into_the_response(
+            self, api_v3_client, plugin_dir):
+        (plugin_dir / "authenticate_spotify.py").write_text(
+            "def load_spotify_credentials():\n"
+            "    return ('id-abc', 'super-secret-value', None)\n"
+        )
+        response = api_v3_client.post(self.URL, json={})
+        assert "super-secret-value" not in response.get_data(as_text=True)
+
+    def test_script_raising_on_import_is_handled(self, api_v3_client, plugin_dir):
+        (plugin_dir / "authenticate_spotify.py").write_text("raise RuntimeError('boom')\n")
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 500
+        assert response.get_json()["status"] == "error"
+
+    def test_bodyless_post_reaches_step_one(self, api_v3_client, plugin_dir):
+        # Covered by the silent=True fix: previously a 500 from body parsing.
+        response = api_v3_client.post(self.URL)
+        assert response.status_code in (400, 500)
+        assert response.get_json()["status"] == "error"
+
+    def test_whitespace_redirect_url_is_treated_as_absent(
+            self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed()) as run:
+            api_v3_client.post(self.URL, json={"redirect_url": "   "})
+        # Step 2 never runs, so no wrapper is executed.
+        run.assert_not_called()
+
+
+class TestYouTubeMusic:
+    """No wrapper script and no redirect_url — deliberately not symmetric."""
+
+    URL = "/api/v3/plugins/authenticate/ytm"
+
+    def test_missing_plugin_directory_is_404(self, api_v3_client, api_v3_module, tmp_path):
+        api_v3_module.api_v3.plugin_manager.get_plugin_directory.return_value = str(
+            tmp_path / "not-installed")
+        assert api_v3_client.post(self.URL).status_code == 404
+
+    def test_missing_script_is_404(self, api_v3_client, plugin_dir):
+        (plugin_dir / "authenticate_ytm.py").unlink()
+        response = api_v3_client.post(self.URL)
+        assert response.status_code == 404
+        assert "script not found" in response.get_json()["message"]
+
+    def test_success(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed(0, "authorized")):
+            response = api_v3_client.post(self.URL)
+        assert response.status_code == 200
+        assert response.get_json()["output"] == "authorized"
+
+    def test_failure_is_a_400_with_combined_output(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed(1, "out", "err")):
+            response = api_v3_client.post(self.URL)
+        assert response.status_code == 400
+        assert response.get_json()["output"] == "outerr"
+
+    def test_timeout_is_a_408(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run",
+                          side_effect=subprocess.TimeoutExpired("python3", 60)):
+            assert api_v3_client.post(self.URL).status_code == 408
+
+    def test_runs_the_script_directly_without_a_shell(self, api_v3_client, plugin_dir):
+        with patch.object(subprocess, "run", return_value=completed()) as run:
+            api_v3_client.post(self.URL)
+        args, kwargs = run.call_args
+        assert args[0][0] == "python3"
+        assert args[0][1].endswith("authenticate_ytm.py")
+        assert kwargs.get("shell") in (None, False)
+        assert kwargs["timeout"] == 60

test/test_api_v3_non_finite_numbers.py

@@ -0,0 +1,85 @@
+"""Non-finite JSON numbers must be rejected, not raise.
+
+json.loads accepts Infinity/-Infinity/NaN by default (they are not valid JSON,
+but Python's parser emits them) and Flask's get_json passes them straight
+through. int(float('inf')) raises OverflowError, which is neither ValueError
+nor TypeError -- so validation blocks that carefully caught those let it
+through and Flask turned it into a 500.
+
+The damage was not the status code. /config/dim-schedule answered with
+CONFIG_SAVE_FAILED and suggested "Check file permissions on config directory"
+and "Check available disk space" for what was actually an invalid number.
+
+NaN already returned 400 (int(nan) raises ValueError), which is why this only
+showed up for the infinities.
+"""
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from test._api_v3_test_helpers import api_v3_client, api_v3_module  # noqa: F401,E402
+
+
+#: (route, field) that returned 500 before OverflowError was caught. Both
+#: infinity signs are exercised: int() raises OverflowError for either, but
+#: only one of them was in the original report, and a guard that special-cased
+#: the sign would pass a one-sided test.
+NON_FINITE_ROUTES = [
+    ('/api/v3/config/dim-schedule', 'dim_brightness'),
+    ('/api/v3/errors/clear', 'max_age_hours'),
+    ('/api/v3/config/main', 'multiplexing'),
+    ('/api/v3/config/main', 'row_address_type'),
+]
+NON_FINITE_CASES = [
+    (route, '{"%s": %s}' % (field, literal))
+    for route, field in NON_FINITE_ROUTES
+    for literal in ('Infinity', '-Infinity')
+]
+
+
+@pytest.mark.parametrize("route,body", NON_FINITE_CASES)
+def test_infinity_is_a_client_error_not_a_server_error(api_v3_client, route, body):
+    """Exactly 400, not merely "some 4xx".
+
+    Accepting any 4xx would let a 404 pass, so renaming one of these routes
+    would leave the test green while testing nothing -- the failure mode this
+    whole file exists to catch.
+    """
+    response = api_v3_client.post(route, data=body, content_type='application/json')
+    assert response.status_code == 400, (
+        f"{route} with {body} answered {response.status_code}; expected 400"
+    )
+
+
+@pytest.mark.parametrize("route,body", [
+    ('/api/v3/config/dim-schedule', '{"dim_brightness": NaN}'),
+    ('/api/v3/errors/clear', '{"max_age_hours": NaN}'),
+])
+def test_nan_is_also_a_client_error(api_v3_client, route, body):
+    """int(nan) raises ValueError so this path already worked -- pinned so a
+    refactor that narrows the except tuple cannot quietly break it."""
+    response = api_v3_client.post(route, data=body, content_type='application/json')
+    assert response.status_code == 400
+
+
+def test_a_valid_number_is_accepted(api_v3_client, api_v3_module, monkeypatch):
+    """Prove the widened except did not start swallowing ordinary input.
+
+    Asserting "not a 400" would not show that: the mocked save path fails for
+    any input, so the assertion would hold even if validation had rejected the
+    value. Give load_config a real dict and stub the atomic save, and the
+    endpoint reaches its success response -- which only happens if 30 passed
+    validation.
+    """
+    api_v3_module.api_v3.config_manager.load_config.return_value = {}
+    monkeypatch.setattr(api_v3_module, '_save_config_atomic',
+                        lambda *a, **k: (True, ''))
+    response = api_v3_client.post(
+        '/api/v3/config/dim-schedule',
+        data='{"dim_brightness": 30}',
+        content_type='application/json',
+    )
+    assert response.status_code == 200, response.get_data(as_text=True)[:200]

test/test_api_v3_optional_body.py

@@ -0,0 +1,136 @@
+"""
+Regression tests: POST endpoints whose body is optional must accept a
+request that has no body at all.
+
+Six handlers in api_v3 read their body as ``request.get_json() or {}``.
+The ``or {}`` states the intent plainly — every field is optional, so a
+bodyless POST should fall back to defaults. But ``get_json()`` without
+``silent=True`` raises ``UnsupportedMediaType`` when the request carries
+no JSON Content-Type, and it raises *before* ``or {}`` is evaluated. Each
+handler's catch-all then turned that into a 500.
+
+So the natural way to call these endpoints — a POST with no body, which
+is what curl, a fetch() without options, and most HTTP clients send by
+default — failed on every one of them. The shipped UI always sends a JSON
+object, which is why this went unnoticed.
+
+This file covers the endpoints whose bodyless behaviour is not already
+tested in their own suite.
+"""
+
+import re
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from test._api_v3_test_helpers import api_v3_client, api_v3_module  # noqa: F401,E402
+
+
+class TestOnDemandStart:
+    URL = "/api/v3/display/on-demand/start"
+
+    def test_bodyless_post_is_not_a_server_error(self, api_v3_client, api_v3_module):
+        response = api_v3_client.post(self.URL)
+        # The endpoint may still reject the request on its own terms (no
+        # plugin_id, nothing to display); what it must not do is fail with
+        # a 500 raised out of body parsing.
+        assert response.status_code != 500
+
+    def test_json_body_still_works(self, api_v3_client, api_v3_module):
+        assert api_v3_client.post(self.URL, json={}).status_code != 500
+
+
+class TestResetPluginConfig:
+    URL = "/api/v3/plugins/config/reset"
+
+    def test_bodyless_post_is_not_a_server_error(self, api_v3_client, api_v3_module):
+        assert api_v3_client.post(self.URL).status_code != 500
+
+    def test_json_body_still_works(self, api_v3_client, api_v3_module):
+        assert api_v3_client.post(self.URL, json={}).status_code != 500
+
+
+class TestDeleteOfTheDayJson:
+    URL = "/api/v3/plugins/of-the-day/json/delete"
+
+    def test_bodyless_post_is_not_a_server_error(self, api_v3_client, api_v3_module):
+        assert api_v3_client.post(self.URL).status_code != 500
+
+    def test_json_body_still_works(self, api_v3_client, api_v3_module):
+        assert api_v3_client.post(self.URL, json={}).status_code != 500
+
+
+class TestPluginLimits:
+    URL = "/api/v3/plugins/clock/limits"
+
+    def test_bodyless_post_is_not_a_server_error(self, api_v3_client, api_v3_module):
+        assert api_v3_client.post(self.URL).status_code != 500
+
+
+class TestMissingBodyGivesTheDeclaredError:
+    """Handlers that answer "No data provided" must actually be able to.
+
+    A second group of handlers reads `data = request.get_json()` and then
+    guards with `if not data: return 400`. That guard is unreachable for a
+    request with no JSON body, because get_json() raises first — so the
+    caller got a 500 "an error occurred; see logs for details" instead of
+    the 400 the handler plainly intends to send.
+    """
+
+    @pytest.mark.parametrize("url", [
+        "/api/v3/plugins/install",
+        "/api/v3/plugins/install-from-url",
+        "/api/v3/plugins/registry-from-url",
+        "/api/v3/config/raw/main",
+        "/api/v3/config/raw/secrets",
+        "/api/v3/cache/delete",
+    ])
+    def test_bodyless_post_gets_a_400_not_a_500(self, api_v3_client, api_v3_module, url):
+        response = api_v3_client.post(url)
+        assert response.status_code == 400, (
+            f"{url} answered {response.status_code}: "
+            f"{response.get_data(as_text=True)[:200]}")
+
+    @pytest.mark.parametrize("url", [
+        "/api/v3/plugins/install",
+        "/api/v3/config/raw/main",
+    ])
+    def test_malformed_json_gets_a_400_not_a_500(self, api_v3_client, api_v3_module, url):
+        response = api_v3_client.post(
+            url, data="{not json", content_type="application/json")
+        assert response.status_code == 400
+
+
+class TestNoBodyReadContradictsItsOwnGuard:
+    SOURCE = Path(__file__).parent.parent / "web_interface/blueprints/api_v3.py"
+
+    def test_no_or_default_read_is_unguarded(self):
+        """`get_json() or <default>` is a contradiction without silent=True.
+
+        Writing `or {}` declares the body optional; omitting silent=True
+        means the call raises before the default can apply.
+        """
+        offenders = [
+            line.strip() for line in self.SOURCE.read_text().splitlines()
+            if "request.get_json()" in line and " or " in line
+        ]
+        assert offenders == [], (
+            "these reads declare a default but raise before reaching it; "
+            f"use get_json(silent=True): {offenders}")
+
+    def test_no_not_data_guard_is_unreachable(self):
+        """A `if not data:` guard needs a read that can actually return None."""
+        lines = self.SOURCE.read_text().splitlines()
+        offenders = []
+        for i, line in enumerate(lines):
+            if re.search(r"=\s*request\.get_json\(\)\s*$", line):
+                window = "\n".join(lines[i + 1:i + 3])
+                if re.search(r"if\s+(not\s+data\b|data\s+is\s+None)", window):
+                    offenders.append(f"line {i + 1}: {line.strip()}")
+        assert offenders == [], (
+            "these handlers guard on a missing body but raise before the "
+            f"guard runs; use get_json(silent=True): {offenders}")

test/test_api_v3_plugin_install_endpoints.py

@@ -0,0 +1,302 @@
+"""
+Endpoint tests for POST /plugins/install and POST /plugins/install-from-url.
+
+Both were only ever tested at the PluginStoreManager layer, so the route
+logic — the queue-vs-direct branch, schema invalidation, plugin discovery,
+state and history recording — was unexercised.
+
+/plugins/install carries the same install logic twice: once inside the
+operation-queue callback and once in the direct fallback. The paired
+tests below assert both branches produce the same side effects, so the
+duplication cannot quietly drift.
+"""
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from test._api_v3_test_helpers import api_v3_client, api_v3_module  # noqa: F401,E402
+
+INSTALL = "/api/v3/plugins/install"
+FROM_URL = "/api/v3/plugins/install-from-url"
+
+
+@pytest.fixture
+def queued(api_v3_module):
+    """Enable the operation queue and run its callback synchronously."""
+    queue = MagicMock()
+
+    def enqueue(operation_type, plugin_id, operation_callback=None):
+        queue.callback_result = operation_callback(MagicMock())
+        return "op-123"
+
+    queue.enqueue_operation.side_effect = enqueue
+    api_v3_module.api_v3.operation_queue = queue
+    return queue
+
+
+def side_effects(module):
+    """The manager calls a successful install is expected to make."""
+    api = module.api_v3
+    return {
+        "schema_invalidated": api.schema_manager.invalidate_cache.call_args_list,
+        "discovered": api.plugin_manager.discover_plugins.call_count,
+        "loaded": api.plugin_manager.load_plugin.call_args_list,
+        "state_set": api.plugin_state_manager.set_plugin_installed.call_args_list,
+        "history": api.operation_history.record_operation.call_args_list,
+    }
+
+
+class TestInstallValidation:
+    def test_uninitialized_store_manager_is_a_500(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager = None
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert response.status_code == 500
+        assert "not initialized" in response.get_json()["message"]
+
+    def test_missing_plugin_id_is_a_400(self, api_v3_client, api_v3_module):
+        response = api_v3_client.post(INSTALL, json={})
+        assert response.status_code == 400
+        assert "plugin_id required" in response.get_json()["message"]
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.assert_not_called()
+
+    def test_empty_body_is_a_400(self, api_v3_client, api_v3_module):
+        assert api_v3_client.post(INSTALL, json=None).status_code == 400
+
+
+class TestInstallDirectPath:
+    """operation_queue is None — the fallback branch."""
+
+    def test_success(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert response.status_code == 200
+        assert response.get_json()["status"] == "success"
+
+    def test_success_side_effects(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        effects = side_effects(api_v3_module)
+        assert effects["schema_invalidated"] == [(("clock",), {})]
+        assert effects["discovered"] == 1
+        assert effects["loaded"] == [(("clock",), {})]
+        assert effects["state_set"] == [(("clock",), {})]
+        assert effects["history"][0].kwargs["status"] == "success"
+
+    def test_branch_forwarded_to_the_manager(self, api_v3_client, api_v3_module):
+        manager = api_v3_module.api_v3.plugin_store_manager
+        manager.install_plugin.return_value = True
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock", "branch": "dev"})
+        manager.install_plugin.assert_called_once_with("clock", branch="dev")
+
+    def test_branch_named_in_the_message(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock", "branch": "dev"})
+        assert "(branch: dev)" in response.get_json()["message"]
+
+    def test_failure_is_a_500(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = False
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert response.status_code == 500
+        assert "Failed to install" in response.get_json()["message"]
+
+    def test_failure_mentions_missing_registry_entry(self, api_v3_client, api_v3_module):
+        manager = api_v3_module.api_v3.plugin_store_manager
+        manager.install_plugin.return_value = False
+        manager.get_plugin_info.return_value = None
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "ghost"})
+        assert "not found in registry" in response.get_json()["message"]
+
+    def test_failure_omits_registry_note_when_plugin_is_known(
+            self, api_v3_client, api_v3_module):
+        manager = api_v3_module.api_v3.plugin_store_manager
+        manager.install_plugin.return_value = False
+        manager.get_plugin_info.return_value = {"id": "clock"}
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert "not found in registry" not in response.get_json()["message"]
+
+    def test_failure_recorded_in_history(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = False
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        record = api_v3_module.api_v3.operation_history.record_operation.call_args
+        assert record.kwargs["status"] == "failed"
+
+    def test_no_side_effects_on_failure(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = False
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        effects = side_effects(api_v3_module)
+        assert effects["schema_invalidated"] == []
+        assert effects["loaded"] == []
+        assert effects["state_set"] == []
+
+
+class TestInstallQueuedPath:
+    """operation_queue present — the callback branch."""
+
+    def test_returns_an_operation_id(self, api_v3_client, api_v3_module, queued):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert response.status_code == 200
+        assert response.get_json()["data"]["operation_id"] == "op-123"
+
+    def test_message_says_queued(self, api_v3_client, api_v3_module, queued):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert "queued" in response.get_json()["message"]
+
+    def test_callback_success_side_effects(self, api_v3_client, api_v3_module, queued):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        effects = side_effects(api_v3_module)
+        assert effects["schema_invalidated"] == [(("clock",), {})]
+        assert effects["discovered"] == 1
+        assert effects["loaded"] == [(("clock",), {})]
+        assert effects["state_set"] == [(("clock",), {})]
+        assert effects["history"][0].kwargs["status"] == "success"
+
+    def test_callback_reports_success(self, api_v3_client, api_v3_module, queued):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert queued.callback_result["success"] is True
+
+    def test_callback_failure_raises_for_the_queue(self, api_v3_client, api_v3_module, queued):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = False
+        # The callback signals failure by raising, so the queue can mark the
+        # operation failed; the route's catch-all turns it into a 500.
+        response = api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        assert response.status_code == 500
+
+    def test_callback_failure_recorded_in_history(self, api_v3_client, api_v3_module, queued):
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = False
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock"})
+        record = api_v3_module.api_v3.operation_history.record_operation.call_args
+        assert record.kwargs["status"] == "failed"
+
+    def test_branch_forwarded_from_the_callback(self, api_v3_client, api_v3_module, queued):
+        manager = api_v3_module.api_v3.plugin_store_manager
+        manager.install_plugin.return_value = True
+        api_v3_client.post(INSTALL, json={"plugin_id": "clock", "branch": "dev"})
+        manager.install_plugin.assert_called_once_with("clock", branch="dev")
+
+
+class TestInstallPathsAgree:
+    """The queue callback and the direct fallback duplicate the same logic."""
+
+    def _run(self, client, module, install_ok, queue):
+        module.api_v3.plugin_store_manager.install_plugin.return_value = install_ok
+        client.post(INSTALL, json={"plugin_id": "clock", "branch": "dev"})
+        return side_effects(module)
+
+    def test_success_side_effects_match(self, api_v3_client, api_v3_module):
+        direct = self._run(api_v3_client, api_v3_module, True, None)
+
+        # Reset and re-run through the queue.
+        for mock in (api_v3_module.api_v3.schema_manager,
+                     api_v3_module.api_v3.plugin_manager,
+                     api_v3_module.api_v3.plugin_state_manager,
+                     api_v3_module.api_v3.operation_history):
+            mock.reset_mock()
+        queue = MagicMock()
+        queue.enqueue_operation.side_effect = (
+            lambda t, p, operation_callback=None: operation_callback(MagicMock()) and "op")
+        api_v3_module.api_v3.operation_queue = queue
+        queued = self._run(api_v3_client, api_v3_module, True, queue)
+
+        assert direct["schema_invalidated"] == queued["schema_invalidated"]
+        assert direct["discovered"] == queued["discovered"]
+        assert direct["loaded"] == queued["loaded"]
+        assert direct["state_set"] == queued["state_set"]
+        assert (direct["history"][0].kwargs["status"]
+                == queued["history"][0].kwargs["status"])
+        assert (direct["history"][0].kwargs["details"]
+                == queued["history"][0].kwargs["details"])
+
+    def test_only_the_message_wording_differs(self, api_v3_client, api_v3_module):
+        # Characterized: the direct path says "Plugin installed
+        # successfully" while the queue callback says "Plugin clock
+        # installed successfully". Cosmetic, and the queue's text is
+        # internal to the operation record rather than the HTTP response.
+        api_v3_module.api_v3.plugin_store_manager.install_plugin.return_value = True
+        direct = api_v3_client.post(INSTALL, json={"plugin_id": "clock"}).get_json()
+        assert direct["message"] == "Plugin installed successfully"
+
+
+class TestInstallFromUrl:
+    def test_uninitialized_store_manager_is_a_500(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager = None
+        assert api_v3_client.post(FROM_URL, json={"repo_url": "http://x"}).status_code == 500
+
+    def test_missing_repo_url_is_a_400(self, api_v3_client, api_v3_module):
+        response = api_v3_client.post(FROM_URL, json={})
+        assert response.status_code == 400
+        assert "repo_url required" in response.get_json()["message"]
+
+    def test_success(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_from_url.return_value = {
+            "success": True, "plugin_id": "clock", "name": "Clock"}
+        response = api_v3_client.post(FROM_URL, json={"repo_url": "https://github.com/o/r"})
+        assert response.status_code == 200
+        body = response.get_json()
+        assert body["plugin_id"] == "clock"
+        assert body["name"] == "Clock"
+
+    def test_all_optional_arguments_forwarded(self, api_v3_client, api_v3_module):
+        manager = api_v3_module.api_v3.plugin_store_manager
+        manager.install_from_url.return_value = {"success": True, "plugin_id": "clock"}
+        api_v3_client.post(FROM_URL, json={
+            "repo_url": "  https://github.com/o/r  ",
+            "plugin_id": "clock",
+            "plugin_path": "plugins/clock",
+            "branch": "dev",
+        })
+        manager.install_from_url.assert_called_once_with(
+            repo_url="https://github.com/o/r",
+            plugin_id="clock",
+            plugin_path="plugins/clock",
+            branch="dev",
+        )
+
+    def test_success_invalidates_schema_and_loads_plugin(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_from_url.return_value = {
+            "success": True, "plugin_id": "clock"}
+        api_v3_client.post(FROM_URL, json={"repo_url": "http://x"})
+        api_v3_module.api_v3.schema_manager.invalidate_cache.assert_called_once_with("clock")
+        api_v3_module.api_v3.plugin_manager.load_plugin.assert_called_once_with("clock")
+
+    def test_success_without_plugin_id_skips_discovery(self, api_v3_client, api_v3_module):
+        # install_from_url can succeed without naming the plugin; there is
+        # then nothing to invalidate or load.
+        api_v3_module.api_v3.plugin_store_manager.install_from_url.return_value = {
+            "success": True, "plugin_id": None}
+        api_v3_client.post(FROM_URL, json={"repo_url": "http://x"})
+        api_v3_module.api_v3.schema_manager.invalidate_cache.assert_not_called()
+        api_v3_module.api_v3.plugin_manager.load_plugin.assert_not_called()
+
+    def test_branch_from_result_included(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_from_url.return_value = {
+            "success": True, "plugin_id": "clock", "branch": "dev"}
+        body = api_v3_client.post(FROM_URL, json={"repo_url": "http://x"}).get_json()
+        assert body["branch"] == "dev"
+        assert "(branch: dev)" in body["message"]
+
+    def test_failure_reports_the_managers_error(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_from_url.return_value = {
+            "success": False, "error": "repo not found"}
+        response = api_v3_client.post(FROM_URL, json={"repo_url": "http://x"})
+        assert response.status_code == 500
+        assert response.get_json()["message"] == "repo not found"
+
+    def test_failure_without_error_uses_fallback_text(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_from_url.return_value = {
+            "success": False}
+        response = api_v3_client.post(FROM_URL, json={"repo_url": "http://x"})
+        assert "Failed to install plugin from URL" in response.get_json()["message"]
+
+    def test_manager_exception_is_a_500(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.install_from_url.side_effect = (
+            RuntimeError("boom"))
+        assert api_v3_client.post(FROM_URL, json={"repo_url": "http://x"}).status_code == 500

test/test_api_v3_registry_endpoints.py

@@ -0,0 +1,179 @@
+"""
+Endpoint tests for the plugin-registry routes in api_v3:
+POST /plugins/store/refresh and POST /plugins/registry-from-url.
+
+Both reach out to the network through PluginStoreManager (mocked here) and
+had no endpoint-level coverage; registry-from-url in particular takes a
+user-supplied URL and hands it straight to the manager.
+"""
+
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from test._api_v3_test_helpers import api_v3_client, api_v3_module  # noqa: F401,E402
+
+
+class TestRefreshPluginStore:
+    URL = "/api/v3/plugins/store/refresh"
+
+    def test_uninitialized_manager_is_a_500(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager = None
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 500
+        assert "not initialized" in response.get_json()["message"]
+
+    def test_success_reports_plugin_count(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.return_value = {
+            "plugins": [{"id": "a"}, {"id": "b"}, {"id": "c"}]}
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 200
+        assert response.get_json()["plugin_count"] == 3
+
+    def test_forces_a_refresh_rather_than_using_cache(self, api_v3_client, api_v3_module):
+        manager = api_v3_module.api_v3.plugin_store_manager
+        manager.fetch_registry.return_value = {"plugins": []}
+        api_v3_client.post(self.URL, json={})
+        manager.fetch_registry.assert_called_once_with(force_refresh=True)
+
+    def test_empty_registry_reports_zero(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.return_value = {}
+        response = api_v3_client.post(self.URL, json={})
+        assert response.get_json()["plugin_count"] == 0
+
+    def test_no_body_is_accepted(self, api_v3_client, api_v3_module):
+        # Regression: `request.get_json() or {}` says a missing body is
+        # fine, but get_json() raises UnsupportedMediaType before `or {}`
+        # is reached, so a bodyless POST — the natural way to call a
+        # refresh endpoint — came back 500.
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.return_value = {"plugins": []}
+        assert api_v3_client.post(self.URL).status_code == 200
+
+    def test_body_without_json_content_type_is_accepted(
+            self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.return_value = {"plugins": []}
+        response = api_v3_client.post(self.URL, data="", content_type="text/plain")
+        assert response.status_code == 200
+
+    def test_malformed_json_body_falls_back_to_defaults(
+            self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.return_value = {"plugins": []}
+        response = api_v3_client.post(
+            self.URL, data="{not json", content_type="application/json")
+        assert response.status_code == 200
+
+    @pytest.mark.parametrize("key", ["fetch_commit_info", "fetch_latest_versions"])
+    def test_either_commit_info_key_extends_the_message(
+            self, api_v3_client, api_v3_module, key):
+        # fetch_latest_versions is the older spelling; both must work.
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.return_value = {"plugins": []}
+        response = api_v3_client.post(self.URL, json={key: True})
+        assert "commit metadata" in response.get_json()["message"]
+
+    def test_message_stays_plain_without_the_flag(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.return_value = {"plugins": []}
+        response = api_v3_client.post(self.URL, json={})
+        assert response.get_json()["message"] == "Plugin store refreshed"
+
+    def test_network_failure_is_a_500(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.side_effect = (
+            ConnectionError("github unreachable"))
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 500
+        assert response.get_json()["message"] == "An error occurred; see logs for details"
+
+    def test_failure_body_carries_no_traceback_or_paths(
+            self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry.side_effect = (
+            RuntimeError("failed at /home/user/LEDMatrix/src/secret.py line 42"))
+        body = api_v3_client.post(self.URL, json={}).get_json()
+        assert "Traceback" not in str(body)
+        # `details` is describe_exception output: one line, type-named,
+        # credential-redacted. It may quote the message, but never a stack.
+        assert body["details"].startswith("RuntimeError:")
+        assert "\n" not in body["details"]
+
+
+class TestRegistryFromUrl:
+    URL = "/api/v3/plugins/registry-from-url"
+
+    def test_uninitialized_manager_is_a_500(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager = None
+        response = api_v3_client.post(self.URL, json={"repo_url": "http://x"})
+        assert response.status_code == 500
+
+    def test_missing_repo_url_is_a_400(self, api_v3_client, api_v3_module):
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 400
+        assert "repo_url required" in response.get_json()["message"]
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.assert_not_called()
+
+    def test_success_returns_the_plugin_list(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.return_value = {
+            "plugins": [{"id": "clock"}]}
+        response = api_v3_client.post(
+            self.URL, json={"repo_url": "https://github.com/o/r"})
+        assert response.status_code == 200
+        body = response.get_json()
+        assert body["plugins"] == [{"id": "clock"}]
+        assert body["registry_url"] == "https://github.com/o/r"
+
+    def test_url_is_trimmed_before_use(self, api_v3_client, api_v3_module):
+        manager = api_v3_module.api_v3.plugin_store_manager
+        manager.fetch_registry_from_url.return_value = {"plugins": []}
+        api_v3_client.post(self.URL, json={"repo_url": "  https://github.com/o/r  "})
+        manager.fetch_registry_from_url.assert_called_once_with("https://github.com/o/r")
+
+    def test_registry_without_plugins_key_returns_empty_list(
+            self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.return_value = {
+            "other": 1}
+        response = api_v3_client.post(self.URL, json={"repo_url": "http://x"})
+        assert response.get_json()["plugins"] == []
+
+    def test_no_registry_found_is_a_400(self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.return_value = None
+        response = api_v3_client.post(self.URL, json={"repo_url": "http://x/not-a-registry"})
+        assert response.status_code == 400
+        assert "Failed to fetch registry" in response.get_json()["message"]
+
+    @pytest.mark.parametrize("url", [
+        "not a url",
+        "javascript:alert(1)",
+        "file:///etc/passwd",
+        "http://localhost:8080/admin",
+    ])
+    def test_unusable_urls_fail_cleanly(self, api_v3_client, api_v3_module, url):
+        # Characterization: the handler performs no URL validation of its
+        # own — whatever the manager makes of the URL decides the outcome.
+        # What is pinned here is that a rejected URL produces a clean 400
+        # rather than a traceback or a 500.
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.return_value = None
+        response = api_v3_client.post(self.URL, json={"repo_url": url})
+        assert response.status_code == 400
+        assert "Traceback" not in str(response.get_json())
+
+    def test_fetch_exception_is_a_500_without_internals(
+            self, api_v3_client, api_v3_module):
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.side_effect = (
+            ValueError("parse failed in /srv/app/internal.py"))
+        response = api_v3_client.post(self.URL, json={"repo_url": "http://x"})
+        assert response.status_code == 500
+        body = response.get_json()
+        assert body["message"] == "An error occurred; see logs for details"
+        assert "Traceback" not in str(body)
+
+    def test_non_string_repo_url_is_rejected(self, api_v3_client, api_v3_module):
+        # Regression: .strip() on a non-string raised, and the catch-all
+        # reported the caller's own mistake as a server fault.
+        response = api_v3_client.post(self.URL, json={"repo_url": 12345})
+        assert response.status_code == 400
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.assert_not_called()
+
+    def test_blank_repo_url_is_rejected(self, api_v3_client, api_v3_module):
+        response = api_v3_client.post(self.URL, json={"repo_url": "   "})
+        assert response.status_code == 400
+        api_v3_module.api_v3.plugin_store_manager.fetch_registry_from_url.assert_not_called()

test/test_api_v3_wifi_endpoints.py

@@ -0,0 +1,240 @@
+"""
+Endpoint tests for the /wifi/* routes in api_v3.
+
+These routes drive the host's actual networking — connecting, dropping a
+connection, switching the radio off — and had no endpoint-level tests at
+all. WiFiManager is mocked throughout; nothing here may touch real
+networking.
+
+Each handler does `from src.wifi_manager import WiFiManager` inside the
+function body, so the patch target is the class at its definition site.
+"""
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from test._api_v3_test_helpers import api_v3_client, api_v3_module  # noqa: F401,E402
+
+
+@pytest.fixture
+def wifi_manager():
+    """Patch WiFiManager where it is defined; yield the instance mock."""
+    with patch("src.wifi_manager.WiFiManager") as cls:
+        instance = MagicMock()
+        cls.return_value = instance
+        yield instance
+
+
+class TestConnect:
+    URL = "/api/v3/wifi/connect"
+
+    def test_success(self, api_v3_client, wifi_manager):
+        wifi_manager.connect_to_network.return_value = (True, "Connected to HomeNet")
+        response = api_v3_client.post(self.URL, json={"ssid": "HomeNet", "password": "pw"})
+        assert response.status_code == 200
+        assert response.get_json()["message"] == "Connected to HomeNet"
+        wifi_manager.connect_to_network.assert_called_once_with("HomeNet", "pw")
+
+    def test_missing_body_rejected(self, api_v3_client, wifi_manager):
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 400
+        wifi_manager.connect_to_network.assert_not_called()
+
+    def test_missing_ssid_rejected(self, api_v3_client, wifi_manager):
+        response = api_v3_client.post(self.URL, json={"password": "pw"})
+        assert response.status_code == 400
+        assert "SSID is required" in response.get_json()["message"]
+        wifi_manager.connect_to_network.assert_not_called()
+
+    @pytest.mark.parametrize("ssid", ["", "   ", "\t"])
+    def test_blank_ssid_rejected(self, api_v3_client, wifi_manager, ssid):
+        response = api_v3_client.post(self.URL, json={"ssid": ssid})
+        assert response.status_code == 400
+        wifi_manager.connect_to_network.assert_not_called()
+
+    def test_ssid_is_trimmed(self, api_v3_client, wifi_manager):
+        wifi_manager.connect_to_network.return_value = (True, "ok")
+        api_v3_client.post(self.URL, json={"ssid": "  HomeNet  "})
+        wifi_manager.connect_to_network.assert_called_once_with("HomeNet", "")
+
+    def test_missing_password_becomes_empty_string(self, api_v3_client, wifi_manager):
+        wifi_manager.connect_to_network.return_value = (True, "ok")
+        api_v3_client.post(self.URL, json={"ssid": "OpenNet"})
+        wifi_manager.connect_to_network.assert_called_once_with("OpenNet", "")
+
+    def test_null_password_becomes_empty_string(self, api_v3_client, wifi_manager):
+        wifi_manager.connect_to_network.return_value = (True, "ok")
+        api_v3_client.post(self.URL, json={"ssid": "OpenNet", "password": None})
+        wifi_manager.connect_to_network.assert_called_once_with("OpenNet", "")
+
+    def test_failure_reports_the_managers_reason(self, api_v3_client, wifi_manager):
+        wifi_manager.connect_to_network.return_value = (False, "Bad password")
+        response = api_v3_client.post(self.URL, json={"ssid": "HomeNet"})
+        assert response.status_code == 400
+        assert response.get_json()["message"] == "Bad password"
+
+    def test_failure_without_reason_uses_fallback_text(self, api_v3_client, wifi_manager):
+        wifi_manager.connect_to_network.return_value = (False, None)
+        response = api_v3_client.post(self.URL, json={"ssid": "HomeNet"})
+        assert response.status_code == 400
+        assert response.get_json()["message"] == "Failed to connect to network"
+
+    def test_manager_exception_is_a_500_without_leaking_internals(
+            self, api_v3_client, wifi_manager):
+        wifi_manager.connect_to_network.side_effect = RuntimeError(
+            "/usr/lib/secret/path blew up")
+        response = api_v3_client.post(self.URL, json={"ssid": "HomeNet"})
+        assert response.status_code == 500
+        body = response.get_json()
+        assert body["message"] == "An error occurred; see logs for details"
+        # `details` comes from describe_exception, which is deliberately
+        # safe to return (redacted, capped) — it names the type.
+        assert "RuntimeError" in body["details"]
+
+
+class TestDisconnect:
+    URL = "/api/v3/wifi/disconnect"
+
+    def test_success(self, api_v3_client, wifi_manager):
+        wifi_manager.disconnect_from_network.return_value = (True, "Disconnected")
+        response = api_v3_client.post(self.URL)
+        assert response.status_code == 200
+        assert response.get_json()["message"] == "Disconnected"
+
+    def test_failure(self, api_v3_client, wifi_manager):
+        wifi_manager.disconnect_from_network.return_value = (False, "Not connected")
+        response = api_v3_client.post(self.URL)
+        assert response.status_code == 400
+        assert response.get_json()["message"] == "Not connected"
+
+    def test_failure_without_reason_uses_fallback(self, api_v3_client, wifi_manager):
+        wifi_manager.disconnect_from_network.return_value = (False, "")
+        response = api_v3_client.post(self.URL)
+        assert response.get_json()["message"] == "Failed to disconnect from network"
+
+    def test_exception_is_a_500(self, api_v3_client, wifi_manager):
+        wifi_manager.disconnect_from_network.side_effect = OSError("nmcli missing")
+        assert api_v3_client.post(self.URL).status_code == 500
+
+
+class TestApMode:
+    ENABLE = "/api/v3/wifi/ap/enable"
+    DISABLE = "/api/v3/wifi/ap/disable"
+
+    def test_enable_success(self, api_v3_client, wifi_manager):
+        wifi_manager.enable_ap_mode.return_value = (True, "AP enabled")
+        response = api_v3_client.post(self.ENABLE, json={})
+        assert response.status_code == 200
+        wifi_manager.enable_ap_mode.assert_called_once_with(force=False)
+
+    @pytest.mark.parametrize("raw,expected", [
+        (True, True), (False, False),
+        ("true", True), ("TRUE", True), ("1", True),
+        ("false", False), ("no", False), ("yes", False),
+        (1, False),  # only real True or the listed strings count
+    ])
+    def test_force_coercion(self, api_v3_client, wifi_manager, raw, expected):
+        wifi_manager.enable_ap_mode.return_value = (True, "ok")
+        api_v3_client.post(self.ENABLE, json={"force": raw})
+        wifi_manager.enable_ap_mode.assert_called_once_with(force=expected)
+
+    def test_enable_without_body(self, api_v3_client, wifi_manager):
+        wifi_manager.enable_ap_mode.return_value = (True, "ok")
+        assert api_v3_client.post(self.ENABLE).status_code == 200
+
+    def test_enable_failure(self, api_v3_client, wifi_manager):
+        wifi_manager.enable_ap_mode.return_value = (False, "hostapd missing")
+        response = api_v3_client.post(self.ENABLE, json={})
+        assert response.status_code == 400
+        assert response.get_json()["message"] == "hostapd missing"
+
+    def test_disable_success(self, api_v3_client, wifi_manager):
+        wifi_manager.disable_ap_mode.return_value = (True, "AP disabled")
+        assert api_v3_client.post(self.DISABLE).status_code == 200
+
+    def test_disable_failure(self, api_v3_client, wifi_manager):
+        wifi_manager.disable_ap_mode.return_value = (False, "not running")
+        assert api_v3_client.post(self.DISABLE).status_code == 400
+
+    def test_enable_exception_is_a_500(self, api_v3_client, wifi_manager):
+        wifi_manager.enable_ap_mode.side_effect = RuntimeError("boom")
+        assert api_v3_client.post(self.ENABLE, json={}).status_code == 500
+
+
+class TestRadio:
+    URL = "/api/v3/wifi/radio"
+
+    def test_get_state(self, api_v3_client, wifi_manager):
+        wifi_manager.get_wifi_radio_state.return_value = {
+            "enabled": True, "ethernet_connected": False}
+        response = api_v3_client.get(self.URL)
+        assert response.status_code == 200
+        assert response.get_json()["data"]["enabled"] is True
+
+    def test_get_state_exception_is_a_500(self, api_v3_client, wifi_manager):
+        wifi_manager.get_wifi_radio_state.side_effect = OSError("rfkill missing")
+        assert api_v3_client.get(self.URL).status_code == 500
+
+    def test_enabled_is_required(self, api_v3_client, wifi_manager):
+        response = api_v3_client.post(self.URL, json={})
+        assert response.status_code == 400
+        assert "enabled is required" in response.get_json()["message"]
+        wifi_manager.set_wifi_radio.assert_not_called()
+
+    def test_enable_success(self, api_v3_client, wifi_manager):
+        wifi_manager.set_wifi_radio.return_value = (True, "Radio on", None)
+        wifi_manager.get_wifi_radio_state.return_value = {"enabled": True}
+        response = api_v3_client.post(self.URL, json={"enabled": True})
+        assert response.status_code == 200
+        wifi_manager.set_wifi_radio.assert_called_once_with(True, force=False)
+
+    @pytest.mark.parametrize("raw,expected", [
+        (True, True), ("true", True), ("1", True), ("yes", True),
+        (False, False), ("false", False), ("off", False), (0, False),
+    ])
+    def test_enabled_coercion_is_string_aware(
+            self, api_v3_client, wifi_manager, raw, expected):
+        # bool("false") is True, so the endpoint parses strings explicitly
+        # rather than trusting truthiness — it is a public contract, not
+        # only the shipped UI which always sends real JSON booleans.
+        wifi_manager.set_wifi_radio.return_value = (True, "ok", None)
+        wifi_manager.get_wifi_radio_state.return_value = {}
+        api_v3_client.post(self.URL, json={"enabled": raw})
+        wifi_manager.set_wifi_radio.assert_called_once_with(expected, force=False)
+
+    def test_force_passed_through(self, api_v3_client, wifi_manager):
+        wifi_manager.set_wifi_radio.return_value = (True, "ok", None)
+        wifi_manager.get_wifi_radio_state.return_value = {}
+        api_v3_client.post(self.URL, json={"enabled": False, "force": "true"})
+        wifi_manager.set_wifi_radio.assert_called_once_with(False, force=True)
+
+    def test_refusal_reports_reason(self, api_v3_client, wifi_manager):
+        # Disabling the radio without Ethernet would lock the user out of
+        # this very interface, so the manager can refuse with a reason.
+        wifi_manager.set_wifi_radio.return_value = (
+            False, "Refusing: no wired fallback", "no_ethernet")
+        response = api_v3_client.post(self.URL, json={"enabled": False})
+        assert response.status_code == 400
+        body = response.get_json()
+        assert body["reason"] == "no_ethernet"
+        assert "Refusing" in body["message"]
+
+    def test_exception_is_a_500(self, api_v3_client, wifi_manager):
+        wifi_manager.set_wifi_radio.side_effect = RuntimeError("boom")
+        assert api_v3_client.post(self.URL, json={"enabled": True}).status_code == 500
+
+
+class TestNoRealNetworking:
+    def test_wifi_manager_is_never_constructed_for_real(self, api_v3_client):
+        # Guard against a future refactor moving the import to module level,
+        # where the fixture's patch of the definition site would stop
+        # applying and the tests would start driving real networking.
+        with patch("src.wifi_manager.WiFiManager") as cls:
+            cls.return_value.disconnect_from_network.return_value = (True, "ok")
+            api_v3_client.post("/api/v3/wifi/disconnect")
+            assert cls.called

test/test_app_plugins_dir_resolution.py

@@ -0,0 +1,52 @@
+"""web_interface/app.py resolves plugins_dir from config at import time.
+
+`project_root` used to be assigned only inside the `else` branch (the
+relative-path case). SchemaManager and composer_bp are wired up with
+`project_root` further down the same module, so a device configured with an
+*absolute* `plugin_system.plugins_directory` (a supported value -- see
+CLAUDE.md) hit `UnboundLocalError` importing web_interface.app at all, taking
+the whole web UI down.
+
+app.py isn't imported directly by the test suite -- it has real side effects
+(managers, background state) at import time -- so this extracts the exact
+resolution block by its stable comment markers and executes it in isolation,
+the same way the rest of the suite validates generated/templated code rather
+than re-implementing it.
+"""
+from pathlib import Path
+
+APP_PY = Path(__file__).resolve().parent.parent / "web_interface/app.py"
+
+START_MARKER = "# Resolve plugin directory - handle both absolute and relative paths\n"
+END_MARKER = "\nplugin_manager = PluginManager("
+
+
+def _resolution_block() -> str:
+    src = APP_PY.read_text()
+    start = src.index(START_MARKER)
+    end = src.index(END_MARKER, start)
+    return src[start:end]
+
+
+def test_resolution_block_still_matches_expected_markers():
+    """If app.py is restructured enough that these markers move, the exec
+    below would silently test nothing -- fail loudly instead."""
+    block = _resolution_block()
+    assert "plugins_dir" in block and "os.path.isabs" in block
+
+
+def test_absolute_plugins_directory_does_not_raise_unboundlocalerror():
+    import os
+    ns = {"os": os, "Path": Path, "__file__": str(APP_PY),
+          "plugins_dir_name": "/opt/ledmatrix-plugins"}
+    exec(compile(_resolution_block(), str(APP_PY), "exec"), ns)  # noqa: S102
+    assert ns["project_root"] == APP_PY.parent.parent
+    assert ns["plugins_dir"] == Path("/opt/ledmatrix-plugins")
+
+
+def test_relative_plugins_directory_still_resolves_under_project_root():
+    import os
+    ns = {"os": os, "Path": Path, "__file__": str(APP_PY),
+          "plugins_dir_name": "plugin-repos"}
+    exec(compile(_resolution_block(), str(APP_PY), "exec"), ns)  # noqa: S102
+    assert ns["plugins_dir"] == ns["project_root"] / "plugin-repos"

test/test_background_fetch_dedupe.py

@@ -0,0 +1,450 @@
+"""A second request for a key already being fetched must join, not duplicate.
+
+request_id embeds a millisecond timestamp and active_requests is keyed by it,
+so every submit looked new and nothing compared what was actually being
+fetched. On a real board the season-schedule cache_key is requested by both
+the Recent and the Upcoming manager: they miss the cache in the same
+millisecond and each start a full download and parse of the same payload.
+Measured on a running board, 138 background fetches in 24 hours arriving in
+pairs at identical timestamps -- half of them redundant.
+
+The cost of a duplicate is a second download, a second JSON parse (the
+expensive part on a Pi), and a second parsed copy resident at the same time.
+Schedules on that board run from 256KB to 20MB. It also consumes a second of
+the three executor slots with identical work, which is what makes two large
+parses peak simultaneously.
+"""
+
+import threading
+import time
+from unittest.mock import MagicMock, Mock, patch
+
+import pytest
+import requests
+
+from src.background_data_service import BackgroundDataService, FetchStatus
+
+
+PAYLOAD = {"events": [{"id": f"g{i}"} for i in range(20)]}
+
+
+@pytest.fixture
+def cache():
+    m = MagicMock()
+    m.get.return_value = None          # always a miss: force the fetch path
+    m.set.return_value = None
+    return m
+
+
+@pytest.fixture
+def service(cache):
+    svc = BackgroundDataService(cache, max_workers=3, request_timeout=5)
+    yield svc
+    svc.shutdown(wait=False)
+
+
+def _resp():
+    r = Mock()
+    r.json.return_value = PAYLOAD
+    r.raise_for_status.return_value = None
+    return r
+
+
+def _wait(service, req_id, timeout=5):
+    deadline = time.time() + timeout
+    while not service.is_request_complete(req_id) and time.time() < deadline:
+        time.sleep(0.02)
+
+
+class _BlockingSession:
+    """Holds the first fetch open so a second can be submitted mid-flight."""
+
+    def __init__(self):
+        self.calls = 0
+        self.release = threading.Event()
+        self.started = threading.Event()
+
+    def get(self, *a, **k):
+        self.calls += 1
+        self.started.set()
+        self.release.wait(timeout=5)
+        return _resp()
+
+
+def test_a_second_submit_for_the_same_key_does_not_fetch_twice(service):
+    session = _BlockingSession()
+    with patch.object(service, "session", session):
+        first = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="nba_2026",
+            callback=lambda r: None, max_retries=0)
+        assert session.started.wait(timeout=5)
+
+        second = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="nba_2026",
+            callback=lambda r: None, max_retries=0)
+
+        assert second == first, "the joiner should share the in-flight request id"
+        session.release.set()
+        _wait(service, first)
+
+    assert session.calls == 1, f"the payload was fetched {session.calls} times"
+
+
+def test_the_joiner_still_gets_its_callback(service):
+    session = _BlockingSession()
+    seen = []
+    with patch.object(service, "session", session):
+        first = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: seen.append("first"), max_retries=0)
+        assert session.started.wait(timeout=5)
+        joined = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: seen.append("second"), max_retries=0)
+        # Assert the coalescing happened, otherwise this passes trivially:
+        # two independent requests would each fire their own callback and the
+        # test would say nothing about the joined path.
+        assert joined == first
+        session.release.set()
+        _wait(service, first)
+
+    deadline = time.time() + 5
+    while len(seen) < 2 and time.time() < deadline:
+        time.sleep(0.02)
+    assert sorted(seen) == ["first", "second"], (
+        f"both submitters must be called back, got {seen}")
+
+
+def test_one_callback_raising_does_not_silence_the_other(service):
+    session = _BlockingSession()
+    seen = []
+
+    def boom(result):
+        raise RuntimeError("consumer blew up")
+
+    with patch.object(service, "session", session):
+        first = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=boom, max_retries=0)
+        assert session.started.wait(timeout=5)
+        joined = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: seen.append("survivor"), max_retries=0)
+        # Same reason: without coalescing these are separate requests and
+        # neither callback can affect the other.
+        assert joined == first
+        session.release.set()
+        _wait(service, first)
+
+    deadline = time.time() + 5
+    while not seen and time.time() < deadline:
+        time.sleep(0.02)
+    assert seen == ["survivor"]
+
+
+def test_different_keys_are_not_coalesced(service):
+    session = _BlockingSession()
+    with patch.object(service, "session", session):
+        a = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/a", cache_key="key_a",
+            callback=lambda r: None, max_retries=0)
+        assert session.started.wait(timeout=5)
+        b = service.submit_fetch_request(
+            sport="nhl", year=2026, url="https://x/b", cache_key="key_b",
+            callback=lambda r: None, max_retries=0)
+        assert a != b, "different cache keys must not share a request"
+        session.release.set()
+        _wait(service, a)
+        _wait(service, b)
+    assert session.calls == 2
+
+
+def test_a_later_submit_after_completion_fetches_again(service):
+    """Dedupe is for concurrent requests only, not a second cache layer."""
+    with patch.object(service.session, "get", side_effect=[_resp(), _resp()]) as get:
+        first = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: None, max_retries=0)
+        _wait(service, first)
+        second = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: None, max_retries=0)
+        _wait(service, second)
+    assert first != second
+    assert get.call_count == 2
+
+
+def test_cancelling_releases_the_key(service):
+    """A cancelled request must not wedge its key against future fetches."""
+    session = _BlockingSession()
+    with patch.object(service, "session", session):
+        first = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: None, max_retries=0)
+        assert session.started.wait(timeout=5)
+        service.cancel_request(first)
+        assert "k" not in service._inflight_by_cache_key
+        session.release.set()
+
+
+def test_a_stranded_index_entry_cannot_wedge_a_key(service):
+    """Defensive: the request is looked up, not trusted from the id alone."""
+    service._inflight_by_cache_key["ghost"] = "no_such_request"
+    with patch.object(service.session, "get", return_value=_resp()):
+        req = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="ghost",
+            callback=lambda r: None, max_retries=0)
+        _wait(service, req)
+    assert service.get_result(req).success is True
+
+
+def test_the_deduplicated_count_is_reported(service):
+    session = _BlockingSession()
+    with patch.object(service, "session", session):
+        first = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: None, max_retries=0)
+        assert session.started.wait(timeout=5)
+        service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            max_retries=0)
+        session.release.set()
+        _wait(service, first)
+    assert service.get_statistics().get("deduplicated_requests") == 1
+
+
+def test_a_cancelled_worker_cannot_overwrite_its_replacement(service, cache):
+    """Cancelling frees the key, so a replacement may already own it.
+
+    The worker cannot abort an HTTP call in flight, so when the cancelled one
+    finally returns it must discard its response rather than write it. Without
+    that, the sequence is: cancel A, submit B for the same key, B fetches and
+    caches fresh data, A returns and overwrites it with the response nobody
+    wanted -- and calls A's callbacks too.
+    """
+    slow = _BlockingSession()
+    stale = {"events": [{"id": "STALE"}]}
+    slow_resp = Mock()
+    slow_resp.json.return_value = stale
+    slow_resp.raise_for_status.return_value = None
+
+    def blocked_get(*a, **k):
+        slow.calls += 1
+        slow.started.set()
+        slow.release.wait(timeout=5)
+        return slow_resp
+
+    called = []
+    with patch.object(service.session, "get", side_effect=blocked_get):
+        first = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: called.append("cancelled_one"), max_retries=0)
+        assert slow.started.wait(timeout=5)
+
+        service.cancel_request(first)
+        assert "k" not in service._inflight_by_cache_key
+
+    # The replacement writes the fresh value while the cancelled fetch is held.
+    fresh = {"events": [{"id": "FRESH"}]}
+    fresh_resp = Mock()
+    fresh_resp.json.return_value = fresh
+    fresh_resp.raise_for_status.return_value = None
+    with patch.object(service.session, "get", return_value=fresh_resp):
+        second = service.submit_fetch_request(
+            sport="nba", year=2026, url="https://x/s", cache_key="k",
+            callback=lambda r: called.append("replacement"), max_retries=0)
+        _wait(service, second)
+
+    assert cache.set.call_args[0][1] == fresh, "replacement must own the cache"
+
+    # Now let the cancelled fetch finish. It must write nothing and call nobody.
+    # Wait for the worker to actually finish rather than sleeping: a fixed
+    # sleep is a race under load, and a slow worker would make this pass for
+    # the wrong reason. A cancelled request is still filed in
+    # completed_requests, so that is the signal it has run to completion.
+    writes_before = cache.set.call_count
+    slow.release.set()
+    deadline = time.time() + 5
+    while first not in service.completed_requests and time.time() < deadline:
+        time.sleep(0.02)
+    assert first in service.completed_requests, "cancelled worker never finished"
+
+    assert cache.set.call_count == writes_before, (
+        "the cancelled worker wrote to the cache after its replacement")
+    assert cache.set.call_args[0][1] == fresh, "stale data overwrote fresh"
+    assert "cancelled_one" not in called, (
+        "a cancelled request must not deliver callbacks")
+
+
+def test_request_ids_are_unique_within_a_millisecond(service):
+    """request_id was sport_year_milliseconds, which collides.
+
+    Two submits inside the same millisecond produced the SAME id, so one
+    silently replaced the other in active_requests and completed_requests.
+    Dedupe hands this id back to every joiner as their handle for
+    get_result(), so uniqueness is now load-bearing rather than incidental.
+    """
+    # Stub the executor rather than the session: this is about what submit
+    # hands back, and letting 50 workers loose would outlive the patch and
+    # make real network calls.
+    with patch.object(service.executor, "submit"):
+        ids = [
+            service.submit_fetch_request(
+                sport="nba", year=2026, url="https://x/s",
+                cache_key=f"key_{i}",           # distinct keys: no dedupe
+                callback=lambda r: None, max_retries=0)
+            for i in range(50)
+        ]
+    assert len(set(ids)) == len(ids), "request ids collided"
+
+
+# --- cancellation must be terminal ------------------------------------------
+#
+# Cancelling used to be advisory: three separate paths wrote request.status
+# without checking whether the request had already been cancelled, so a cancel
+# could be silently undone and the work it was meant to stop went ahead.
+
+URL = "http://example.invalid/scores"
+KEY = "sched_nfl_2025"
+
+
+class _CountingSession:
+    """Records whether an HTTP fetch was ever attempted."""
+
+    def __init__(self):
+        self.calls = 0
+
+    def get(self, *a, **k):
+        self.calls += 1
+        return _resp()
+
+
+class _BlockingFailingSession(_CountingSession):
+    """Holds the fetch open, then fails it.
+
+    Cancelling while the worker is parked inside the HTTP call is the only way
+    to reach the exception handler as a cancelled request. Cancel it before
+    the call starts and the worker returns at the pre-start branch instead,
+    which would leave the except path untested.
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.started = threading.Event()
+        self.release = threading.Event()
+
+    def get(self, *a, **k):
+        self.calls += 1
+        self.started.set()
+        self.release.wait(timeout=5)
+        raise requests.RequestException("connection reset")
+
+
+def _fill_every_worker_slot(service, slots=3):
+    """Occupy the pool so the next submit is queued rather than started.
+
+    This is what makes "cancel before the worker runs" deterministic instead
+    of a race the test would win only sometimes. Returns the gate that
+    releases the pool.
+    """
+    gate = threading.Event()
+    for _ in range(slots):
+        service.executor.submit(gate.wait, 5)
+    return gate
+
+
+def test_cancelling_before_the_worker_starts_stops_the_fetch(service, cache):
+    """The queued worker must honour a cancel, not overwrite it with IN_PROGRESS.
+
+    Between submit and the worker picking the job up, the request sits in the
+    executor queue. Cancelling there is the cheapest possible cancel -- nothing
+    has been downloaded yet -- and it was the one that did not work.
+    """
+    gate = _fill_every_worker_slot(service)
+    session = _CountingSession()
+    delivered = []
+
+    with patch.object(service, 'session', session):
+        rid = service.submit_fetch_request(
+            "nfl", 2025, URL, KEY, max_retries=0, callback=delivered.append
+        )
+        assert service.cancel_request(rid) is True
+        gate.set()                       # let the queued worker run
+        _wait(service, rid)
+
+    assert session.calls == 0, (
+        "cancelled before it started, yet the worker still downloaded the payload"
+    )
+    assert cache.set.call_count == 0, "a cancelled request wrote to the cache"
+    assert delivered == [], "a cancelled request invoked its callbacks"
+
+
+def test_a_cancel_during_the_commit_is_refused(service, cache):
+    """Once the worker has claimed the commit, cancelling is too late.
+
+    The claim and the cancelled-check happen in one critical section, so a
+    cancel arriving after it cannot retroactively abandon data already on its
+    way to the cache. Letting it through stranded every joiner: the payload
+    landed in the cache but the callbacks were suppressed, so a manager that
+    joined this fetch waited for a call that never came.
+    """
+    gate = _fill_every_worker_slot(service)
+    late = {}
+    delivered = []
+
+    def cancel_mid_write(key, data, *a, **k):
+        late['returned'] = service.cancel_request(late['rid'])
+
+    cache.set.side_effect = cancel_mid_write
+
+    # Read the payload inside the callback. The service releases result.data
+    # once every callback has been delivered, so inspecting the FetchResult
+    # afterwards sees the released object, not what the caller was handed.
+    def record(result):
+        delivered.append((result.success, result.data))
+
+    with patch.object(service, 'session', _CountingSession()):
+        late['rid'] = service.submit_fetch_request(
+            "nfl", 2025, URL, KEY, max_retries=0, callback=record
+        )
+        gate.set()                       # only now can the worker reach the commit
+        _wait(service, late['rid'])
+
+    assert late.get('returned') is False, (
+        "cancelled a request that had already committed"
+    )
+    assert cache.set.call_count == 1, "the commit itself was lost"
+    assert delivered == [(True, PAYLOAD)], (
+        "data reached the cache but the callbacks were suppressed -- "
+        "every joined submitter is left waiting forever"
+    )
+
+
+def test_a_failure_after_cancelling_stays_cancelled(service, cache):
+    """A cancelled request that then errors must not resurface as FAILED.
+
+    The except path overwrote CANCELLED with FAILED, and the callback gate in
+    the finally block only suppresses callbacks for CANCELLED -- so cancelling
+    a request that was about to time out delivered a spurious error callback.
+    """
+    session = _BlockingFailingSession()
+    delivered = []
+
+    with patch.object(service, 'session', session):
+        rid = service.submit_fetch_request(
+            "nfl", 2025, URL, KEY, max_retries=0, callback=delivered.append
+        )
+        # Assert the worker is inside the HTTP call before cancelling,
+        # otherwise this silently degrades into the pre-start case and the
+        # exception handler is never exercised.
+        assert session.started.wait(timeout=5)
+        assert service.cancel_request(rid) is True
+        session.release.set()
+        _wait(service, rid)
+
+    assert session.calls == 1, "the fetch never started, so nothing could fail"
+
+    assert delivered == [], "a cancelled request delivered a failure callback"
+    assert service.get_request_status(rid) is FetchStatus.CANCELLED, (
+        "a cancelled request that then errored was reported as FAILED"
+    )

test/test_background_payload_release.py

@@ -0,0 +1,280 @@
+"""A delivered fetch payload must not stay resident on the stored result.
+
+BackgroundDataService kept the fetched body on the FetchResult it filed in
+`completed_requests`, which is swept only hourly and capped at 500 entries by
+count. For status records that is free; for a season schedule it is not. NCAA
+football's 2026 schedule is 946 games, and on a 1GB Pi 3B+ the parsed payload
+measured ~90MB -- a tenth of the board's memory, pinned for an hour after the
+consumer had already been handed it.
+
+The cache-hit path was the worse of the two. It runs once per update interval
+per sport, mints a fresh request_id each time, and hands back whatever the
+cache returns -- so a memory-tier miss (the tier is capped at 150 entries)
+re-parses the payload from disk into a genuinely new object. Those accumulate
+as separate copies rather than shared references, which is the staircase seen
+in the field: RSS stepping up ~90MB per sport as seasons loaded and never
+coming back down.
+
+Releasing is safe because the payload is written to the cache under the
+request's cache_key before the result is built, and that is where consumers
+read it from -- the callback is handed the object directly and the plugins use
+it only in passing before reading the cache back.
+
+Requests submitted *without* a callback keep their payload: polling
+get_result() is then the only way to collect it, so releasing would break that
+contract.
+
+And the release must happen after EVERY callback, not after each one. Callers
+that joined an in-flight fetch share a single FetchResult, so releasing per
+delivery strips the payload out from under everyone still queued -- see
+TestJoinersAllGetTheData.
+"""
+
+import threading
+import time
+import pytest
+from unittest.mock import MagicMock, Mock, patch
+
+from src.background_data_service import BackgroundDataService
+
+
+PAYLOAD = {"events": [{"id": f"g{i}"} for i in range(50)]}
+
+
+@pytest.fixture
+def cache():
+    m = MagicMock()
+    m.get.return_value = None
+    m.set.return_value = None
+    return m
+
+
+@pytest.fixture
+def service(cache):
+    svc = BackgroundDataService(cache, max_workers=2, request_timeout=5)
+    yield svc
+    svc.shutdown(wait=False)
+
+
+def _wait(service, req_id, timeout=5):
+    """Wait for the result to be FILED.
+
+    Enough for anything that is true by the time the worker stores the result:
+    its success flag, its error, the cache write that happened during the
+    fetch.
+    """
+    deadline = time.time() + timeout
+    while not service.is_request_complete(req_id) and time.time() < deadline:
+        time.sleep(0.02)
+
+
+def _wait_for_release(service, req_id, timeout=5):
+    """Wait for the payload to be RELEASED, which is strictly later.
+
+    The worker files the result, then runs the callback, then releases. So
+    is_request_complete() goes true while the callback still has not run --
+    waiting on it alone leaves a window in which `seen` is empty and the
+    payload is still resident, and the assertions race the worker. It passes
+    in practice only because a one-line callback usually beats the 20ms poll.
+
+    Release happens after the callback returns, so a released payload also
+    means the callback has finished: one wait covers both.
+    """
+    deadline = time.time() + timeout
+    while time.time() < deadline:
+        result = service.get_result(req_id)
+        if result is not None and result.data is None:
+            return
+        time.sleep(0.02)
+    raise AssertionError(
+        f"payload for {req_id} was never released (callback may not have run)")
+
+
+def _resp():
+    r = Mock()
+    r.json.return_value = PAYLOAD
+    r.raise_for_status.return_value = None
+    return r
+
+
+class TestFetchPath:
+    def test_callback_receives_the_payload_then_it_is_released(self, service, cache):
+        seen = {}
+
+        def callback(result):
+            # The consumer's one look at the data happens here.
+            seen['events'] = len(result.data['events'])
+
+        with patch.object(service.session, "get", return_value=_resp()):
+            req_id = service.submit_fetch_request(
+                sport="ncaa_fb", year=2026, url="https://example.com/s",
+                cache_key="ncaa_fb_2026", callback=callback, max_retries=0,
+            )
+            _wait_for_release(service, req_id)
+
+        assert seen['events'] == 50, "callback must still be handed the payload"
+
+        stored = service.get_result(req_id)
+        assert stored is not None
+        assert stored.success is True
+        assert stored.data is None, "payload must not stay on the stored result"
+
+    def test_nothing_is_lost_the_cache_holds_it(self, service, cache):
+        with patch.object(service.session, "get", return_value=_resp()):
+            req_id = service.submit_fetch_request(
+                sport="ncaa_fb", year=2026, url="https://example.com/s",
+                cache_key="ncaa_fb_2026", callback=lambda r: None, max_retries=0,
+            )
+            _wait(service, req_id)
+
+        cache.set.assert_called_once()
+        key, written = cache.set.call_args[0][:2]
+        assert key == "ncaa_fb_2026"
+        assert written == PAYLOAD, "the payload must be persisted before release"
+
+    def test_without_a_callback_the_payload_is_kept(self, service, cache):
+        # Polling get_result() is then the only delivery mechanism.
+        with patch.object(service.session, "get", return_value=_resp()):
+            req_id = service.submit_fetch_request(
+                sport="nfl", year=2026, url="https://example.com/s",
+                cache_key="nfl_2026", max_retries=0,
+            )
+            _wait(service, req_id)
+
+        assert service.get_result(req_id).data == PAYLOAD
+
+    def test_a_failed_fetch_still_records_its_error(self, service, cache):
+        with patch.object(service.session, "get", side_effect=Exception("boom")):
+            req_id = service.submit_fetch_request(
+                sport="nfl", year=2026, url="https://example.com/s",
+                cache_key="nfl_2026", callback=lambda r: None, max_retries=0,
+            )
+            _wait(service, req_id)
+
+        stored = service.get_result(req_id)
+        assert stored.success is False
+        assert stored.error is not None
+
+
+class TestCacheHitPath:
+    def test_cache_hit_releases_after_the_callback(self, service, cache):
+        cache.get.return_value = PAYLOAD
+        seen = {}
+
+        req_id = service.submit_fetch_request(
+            sport="ncaa_fb", year=2026, url="https://example.com/s",
+            cache_key="ncaa_fb_2026",
+            callback=lambda r: seen.update(events=len(r.data['events'])),
+        )
+
+        assert seen['events'] == 50
+        assert service.get_result(req_id).data is None
+
+    def test_repeated_cache_hits_do_not_accumulate_payloads(self, service, cache):
+        # The staircase: one entry per update interval per sport, each one
+        # potentially a freshly parsed copy after a memory-tier miss.
+        cache.get.return_value = PAYLOAD
+
+        for _ in range(25):
+            service.submit_fetch_request(
+                sport="ncaa_fb", year=2026, url="https://example.com/s",
+                cache_key="ncaa_fb_2026", callback=lambda r: None,
+            )
+
+        retained = [r for r in service.completed_requests.values() if r.data is not None]
+        assert retained == [], f"{len(retained)} payloads still resident"
+
+    def test_cache_hit_without_a_callback_is_unchanged(self, service, cache):
+        cache.get.return_value = PAYLOAD
+        req_id = service.submit_fetch_request(
+            sport="nfl", year=2026, url="https://example.com/s",
+            cache_key="nfl_2026",
+        )
+        assert service.get_result(req_id).data == PAYLOAD
+
+
+class TestJoinersAllGetTheData:
+    """Deduplicated callers share one FetchResult; releasing between them
+    empties it for the rest.
+
+    Not a corner case. A sport's recent, upcoming and live managers all ask for
+    the same season schedule, so the second and third are joiners on almost
+    every cycle. Releasing inside the delivery loop handed the payload to
+    whichever ran first and gave the others `result.data is None`.
+
+    The consequence was worse than a quiet degradation, because consumers do
+    `result.data.get('events')`: they raised AttributeError, the delivery loop
+    caught it, and the whole failure surfaced as a single
+    "Error in callback for request ..." line while that manager silently never
+    received its schedule.
+    """
+
+    class _BlockingSession:
+        """Holds the fetch open so a second submit lands while in flight."""
+
+        def __init__(self):
+            self.release = threading.Event()
+            self.started = threading.Event()
+
+        def get(self, *a, **k):
+            self.started.set()
+            self.release.wait(timeout=5)
+            return _resp()
+
+    def test_every_joiner_is_handed_the_payload(self, service):
+        session = self._BlockingSession()
+        seen = {}
+
+        def record(name):
+            # Read it the way the sport managers do. `result.data['events']`
+            # would raise TypeError on None; `.get` raises AttributeError,
+            # which is the error actually seen in the field.
+            def cb(result):
+                seen[name] = result.data.get('events') if result.data else None
+            return cb
+
+        with patch.object(service, "session", session):
+            first = service.submit_fetch_request(
+                sport="nhl", year=2026, url="https://x/s", cache_key="nhl_2026",
+                callback=record("first"), max_retries=0)
+            assert session.started.wait(timeout=5)
+
+            joined = service.submit_fetch_request(
+                sport="nhl", year=2026, url="https://x/s", cache_key="nhl_2026",
+                callback=record("second"), max_retries=0)
+            # Without coalescing these are two independent fetches that each
+            # own their result, and the test proves nothing about sharing.
+            assert joined == first, "the joiner should share the in-flight id"
+
+            session.release.set()
+            _wait(service, first)
+
+        deadline = time.time() + 5
+        while len(seen) < 2 and time.time() < deadline:
+            time.sleep(0.02)
+
+        assert set(seen) == {"first", "second"}, f"both must be called, got {seen}"
+        for name, events in seen.items():
+            assert events is not None, (
+                f"{name!r} was handed a released payload: the result was "
+                f"emptied before every callback had been delivered")
+            assert len(events) == 50, f"{name!r} got {events!r}"
+
+    def test_the_payload_is_still_released_once_they_have_all_had_it(self, service):
+        """The memory fix must survive the ordering fix."""
+        session = self._BlockingSession()
+
+        with patch.object(service, "session", session):
+            first = service.submit_fetch_request(
+                sport="nhl", year=2026, url="https://x/s", cache_key="nhl_2026",
+                callback=lambda r: None, max_retries=0)
+            assert session.started.wait(timeout=5)
+            service.submit_fetch_request(
+                sport="nhl", year=2026, url="https://x/s", cache_key="nhl_2026",
+                callback=lambda r: None, max_retries=0)
+            session.release.set()
+            _wait_for_release(service, first)
+
+        stored = service.get_result(first)
+        assert stored is not None and stored.data is None, (
+            "the payload must still be dropped once every callback has run")

test/test_base_odds_manager.py

@@ -0,0 +1,365 @@
+"""
+Tests for src/base_odds_manager.py (BaseOddsManager).
+
+Covers get_odds validation/caching/URL construction, the null-safe
+_extract_espn_data fix (ESPN sends explicit JSON nulls for absent sides),
+the no_odds sentinel, stale-cache fallback on request failure,
+is_odds_available's ML-blind truth table, the fixed format_odds_summary
+gate (money-line-only odds now format), get_odds_for_games, and
+configuration loading.
+
+No real network: requests.Session.get is always patched. The odds path sends
+its requests through a session so it can identify itself to ESPN, so patching
+the module-level requests.get would no longer intercept anything.
+"""
+
+from unittest.mock import MagicMock, patch
+
+import pytest
+import requests
+
+from src.base_odds_manager import BaseOddsManager
+
+
+FULL_ITEM = {
+    'details': 'DAL -3.5',
+    'overUnder': 47.5,
+    'spread': -3.5,
+    'homeTeamOdds': {'moneyLine': -150, 'current': {'pointSpread': {'value': -3.5}}},
+    'awayTeamOdds': {'moneyLine': 130, 'current': {'pointSpread': {'value': 3.5}}},
+}
+
+FULL_EXTRACTED = {
+    'details': 'DAL -3.5',
+    'over_under': 47.5,
+    'spread': -3.5,
+    'home_team_odds': {'money_line': -150, 'spread_odds': -3.5},
+    'away_team_odds': {'money_line': 130, 'spread_odds': 3.5},
+}
+
+
+def _make_response(payload):
+    response = MagicMock()
+    response.json.return_value = payload
+    response.raise_for_status.return_value = None
+    return response
+
+
+@pytest.fixture
+def cache_manager():
+    cm = MagicMock()
+    # A bare MagicMock returns truthy Mocks from every call, so every
+    # get_odds() would look like a cache hit. Explicitly wire a miss.
+    cm.get_with_auto_strategy.return_value = None
+    return cm
+
+
+@pytest.fixture
+def manager(cache_manager):
+    return BaseOddsManager(cache_manager)
+
+
+@pytest.fixture
+def mock_get():
+    with patch('src.base_odds_manager.requests.Session.get') as m:
+        m.return_value = _make_response({'items': [dict(FULL_ITEM)]})
+        yield m
+
+
+# ---------------------------------------------------------------------------
+# get_odds
+# ---------------------------------------------------------------------------
+
+class TestGetOdds:
+    def test_none_sport_raises(self, manager):
+        with pytest.raises(ValueError):
+            manager.get_odds(None, 'nfl', '1')
+
+    def test_none_league_raises(self, manager):
+        with pytest.raises(ValueError):
+            manager.get_odds('football', None, '1')
+
+    def test_cache_key_and_url(self, manager, cache_manager, mock_get):
+        manager.get_odds('football', 'nfl', '401')
+
+        cache_manager.get_with_auto_strategy.assert_called_once_with(
+            'odds_espn_football_nfl_401')
+        url = mock_get.call_args[0][0]
+        # Event id appears twice: /events/<id>/competitions/<id>/odds
+        assert '/events/401/competitions/401/odds' in url
+        assert url == ('https://sports.core.api.espn.com/v2/sports/football/'
+                       'leagues/nfl/events/401/competitions/401/odds')
+        # The number matters less than the property: a single stalled request
+        # must not be able to consume the plugin executor's 30s operation
+        # budget, since odds are fetched per live game inside update().
+        assert mock_get.call_args.kwargs['timeout'] == 5
+        assert mock_get.call_args.kwargs['timeout'] < 30
+
+    def test_ncaa_fb_maps_to_college_football(self, manager, mock_get):
+        manager.get_odds('football', 'ncaa_fb', '401')
+
+        url = mock_get.call_args[0][0]
+        assert '/leagues/college-football/' in url
+
+    def test_unknown_league_passes_through(self, manager, mock_get):
+        manager.get_odds('football', 'xfl', '401')
+
+        assert '/leagues/xfl/' in mock_get.call_args[0][0]
+
+    def test_cache_hit_skips_http(self, manager, cache_manager, mock_get):
+        cache_manager.get_with_auto_strategy.return_value = {'spread': -3.0}
+
+        result = manager.get_odds('football', 'nfl', '401')
+
+        assert result == {'spread': -3.0}
+        mock_get.assert_not_called()
+
+    def test_cached_no_odds_sentinel_returned_verbatim(
+            self, manager, cache_manager, mock_get):
+        cache_manager.get_with_auto_strategy.return_value = {'no_odds': True}
+
+        result = manager.get_odds('football', 'nfl', '401')
+
+        assert result == {'no_odds': True}
+        mock_get.assert_not_called()
+        assert manager.is_odds_available(result) is False
+
+    def test_success_caches_extracted_data_with_interval_ttl(
+            self, manager, cache_manager, mock_get):
+        result = manager.get_odds('football', 'nfl', '401',
+                                  update_interval_seconds=100)
+
+        assert result == FULL_EXTRACTED
+        cache_manager.set.assert_called_once_with(
+            'odds_espn_football_nfl_401', FULL_EXTRACTED, ttl=100)
+
+    def test_no_odds_caches_sentinel(self, manager, cache_manager, mock_get):
+        mock_get.return_value = _make_response({'count': 0, 'items': []})
+
+        result = manager.get_odds('football', 'nfl', '401')
+
+        assert result is None
+        cache_manager.set.assert_called_once_with(
+            'odds_espn_football_nfl_401', {'no_odds': True}, ttl=3600)
+
+    def test_zero_interval_falls_back_to_default(
+            self, manager, cache_manager, mock_get):
+        # Quirk pin: `update_interval_seconds or self.update_interval`
+        # treats an explicit 0 as falsy, so the 3600 default wins.
+        manager.get_odds('football', 'nfl', '401', update_interval_seconds=0)
+
+        assert cache_manager.set.call_args.kwargs['ttl'] == 3600
+
+    def test_request_exception_falls_back_to_stale_cache(
+            self, manager, cache_manager, mock_get):
+        cache_manager.get_with_auto_strategy.side_effect = [
+            None, {'stale': True}]
+        mock_get.side_effect = requests.exceptions.RequestException('boom')
+
+        result = manager.get_odds('football', 'nfl', '401')
+
+        assert result == {'stale': True}
+        assert cache_manager.get_with_auto_strategy.call_count == 2
+
+
+# ---------------------------------------------------------------------------
+# _extract_espn_data
+# ---------------------------------------------------------------------------
+
+class TestExtractEspnData:
+    def test_full_item_extracts_all_fields(self, manager):
+        result = manager._extract_espn_data({'items': [dict(FULL_ITEM)]})
+        assert result == FULL_EXTRACTED
+
+    def test_explicit_nulls_do_not_raise(self, manager):
+        # Post-fix: ESPN sends explicit JSON nulls for absent sides
+        # ("homeTeamOdds": null, "current": null); extraction must not
+        # raise and yields None fields.
+        payload = {'items': [{
+            'homeTeamOdds': None,
+            'awayTeamOdds': {'moneyLine': 150, 'current': None},
+        }]}
+
+        result = manager._extract_espn_data(payload)
+
+        assert result is not None
+        assert result['home_team_odds']['money_line'] is None
+        assert result['home_team_odds']['spread_odds'] is None
+        assert result['away_team_odds']['money_line'] == 150
+        assert result['away_team_odds']['spread_odds'] is None
+
+    def test_valid_empty_response_returns_none(self, manager):
+        assert manager._extract_espn_data({'count': 0, 'items': []}) is None
+
+    def test_unexpected_structure_returns_none(self, manager):
+        assert manager._extract_espn_data({'unexpected': True}) is None
+
+    def test_item_without_odds_fields_cached_as_data_not_sentinel(
+            self, manager, cache_manager, mock_get):
+        # Characterization pin: an item with no odds fields still extracts
+        # to a truthy dict of all-None values, so get_odds caches it as
+        # real data (NOT the no_odds sentinel) — but is_odds_available
+        # correctly reports False for it.
+        mock_get.return_value = _make_response({'items': [{}]})
+
+        result = manager.get_odds('football', 'nfl', '401')
+
+        assert result == {
+            'details': None,
+            'over_under': None,
+            'spread': None,
+            'home_team_odds': {'money_line': None, 'spread_odds': None},
+            'away_team_odds': {'money_line': None, 'spread_odds': None},
+        }
+        cache_manager.set.assert_called_once_with(
+            'odds_espn_football_nfl_401', result, ttl=3600)
+        assert manager.is_odds_available(result) is False
+
+
+# ---------------------------------------------------------------------------
+# is_odds_available
+# ---------------------------------------------------------------------------
+
+class TestIsOddsAvailable:
+    def test_none_is_false(self, manager):
+        assert manager.is_odds_available(None) is False
+
+    def test_empty_dict_is_false(self, manager):
+        assert manager.is_odds_available({}) is False
+
+    def test_no_odds_sentinel_is_false(self, manager):
+        assert manager.is_odds_available({'no_odds': True}) is False
+
+    def test_spread_is_true(self, manager):
+        assert manager.is_odds_available({'spread': -3.5}) is True
+
+    def test_over_under_is_true(self, manager):
+        assert manager.is_odds_available({'over_under': 47.5}) is True
+
+    def test_nested_home_spread_odds_is_true(self, manager):
+        assert manager.is_odds_available(
+            {'home_team_odds': {'spread_odds': -3.5}}) is True
+
+    def test_nested_away_spread_odds_is_true(self, manager):
+        assert manager.is_odds_available(
+            {'away_team_odds': {'spread_odds': 3.5}}) is True
+
+    def test_moneyline_only_is_false(self, manager):
+        # Pinned ML-blind contract: is_odds_available ignores money lines
+        # (its callers decide whether to render an odds widget). Note that
+        # format_odds_summary deliberately uses a DIFFERENT gate — it will
+        # still format money-line-only odds (see TestFormatOddsSummary).
+        ml_only = {
+            'home_team_odds': {'money_line': -120},
+            'away_team_odds': {'money_line': 100},
+        }
+        assert manager.is_odds_available(ml_only) is False
+
+
+# ---------------------------------------------------------------------------
+# format_odds_summary (fixed gate: empty / no_odds only)
+# ---------------------------------------------------------------------------
+
+class TestFormatOddsSummary:
+    def test_moneyline_only_formats(self, manager):
+        result = manager.format_odds_summary({
+            'home_team_odds': {'money_line': -120},
+            'away_team_odds': {'money_line': 100},
+        })
+        assert result == 'Home ML: -120 | Away ML: 100'
+
+    def test_full_data_formats_all_parts(self, manager):
+        result = manager.format_odds_summary(FULL_EXTRACTED)
+        assert result == 'Spread: -3.5 | O/U: 47.5 | Home ML: -150 | Away ML: 130'
+
+    def test_none_is_no_odds(self, manager):
+        assert manager.format_odds_summary(None) == 'No odds available'
+
+    def test_empty_dict_is_no_odds(self, manager):
+        assert manager.format_odds_summary({}) == 'No odds available'
+
+    def test_no_odds_sentinel_is_no_odds(self, manager):
+        assert manager.format_odds_summary(
+            {'no_odds': True}) == 'No odds available'
+
+
+# ---------------------------------------------------------------------------
+# get_odds_for_games
+# ---------------------------------------------------------------------------
+
+class TestGetOddsForGames:
+    def test_missing_fields_get_none_odds_without_http(self, manager, mock_get):
+        games = [
+            {'sport': 'football'},
+            {'league': 'nfl'},
+            {'id': '9'},
+            {},
+        ]
+
+        result = manager.get_odds_for_games(games)
+
+        assert all(g['odds'] is None for g in result)
+        mock_get.assert_not_called()
+
+    def test_per_game_exception_continues_loop(self, manager, monkeypatch):
+        def fake_get_odds(sport, league, event_id,
+                          update_interval_seconds=None):
+            if event_id == 'bad':
+                raise RuntimeError('boom')
+            return {'spread': -1.0}
+
+        monkeypatch.setattr(manager, 'get_odds', fake_get_odds)
+        games = [
+            {'sport': 'football', 'league': 'nfl', 'id': 'bad'},
+            {'sport': 'football', 'league': 'nfl', 'id': 'ok'},
+        ]
+
+        result = manager.get_odds_for_games(games)
+
+        assert len(result) == 2
+        assert result[0]['odds'] is None
+        assert result[1]['odds'] == {'spread': -1.0}
+
+    def test_input_dicts_mutated_in_place_and_returned(self, manager, mock_get):
+        # Pin: get_odds_for_games mutates the caller's game dicts in place
+        # and returns the same objects, not copies.
+        game = {'sport': 'football', 'league': 'nfl', 'id': '401'}
+
+        result = manager.get_odds_for_games([game])
+
+        assert result[0] is game
+        assert game['odds'] == FULL_EXTRACTED
+
+
+# ---------------------------------------------------------------------------
+# _load_configuration
+# ---------------------------------------------------------------------------
+
+class TestLoadConfiguration:
+    def test_loads_values_from_config(self, cache_manager):
+        config_manager = MagicMock()
+        config_manager.get_config.return_value = {
+            'base_odds_manager': {
+                'update_interval': 100,
+                'timeout': 5,
+                'cache_ttl': 42,
+            }
+        }
+
+        manager = BaseOddsManager(cache_manager, config_manager=config_manager)
+
+        assert manager.update_interval == 100
+        # Key/attr mismatch pin: the config key is 'timeout' but the
+        # attribute is request_timeout.
+        assert manager.request_timeout == 5
+        assert manager.cache_ttl == 42
+
+    def test_get_config_raising_keeps_defaults(self, cache_manager):
+        config_manager = MagicMock()
+        config_manager.get_config.side_effect = RuntimeError('boom')
+
+        manager = BaseOddsManager(cache_manager, config_manager=config_manager)
+
+        assert manager.update_interval == 3600
+        assert manager.request_timeout == 5
+        assert manager.cache_ttl == 1800

test/test_base_plugin_duration.py

@@ -78,10 +78,20 @@ class TestInstanceVariable:
                              instance_duration=[30])
         assert plugin.get_display_duration() == 20.0
 
-    def test_bool_true_is_one_second(self):
+    def test_bool_true_falls_through_like_any_non_number(self):
-        # Characterized quirk: bool is an int subclass, so display_duration =
+        # bool is an int subclass, but a boolean is not a duration: True
-        # True passes the isinstance((int, float)) branch and returns 1.0.
+        # must NOT read as 1 second — it falls through to config/default.
-        assert make_plugin(instance_duration=True).get_display_duration() == 1.0
+        assert make_plugin(instance_duration=True).get_display_duration() == 15.0
+
+    def test_bool_true_falls_through_to_config(self):
+        plugin = make_plugin(config={"display_duration": 20},
+                             instance_duration=True)
+        assert plugin.get_display_duration() == 20.0
+
+    def test_bool_false_still_falls_through(self):
+        plugin = make_plugin(config={"display_duration": 20},
+                             instance_duration=False)
+        assert plugin.get_display_duration() == 20.0
 
 
 class TestConfigFallback:
@@ -108,3 +118,29 @@ class TestConfigFallback:
 
     def test_config_none_uses_default(self):
         assert make_plugin({"display_duration": None}).get_display_duration() == 15.0
+
+    def test_config_bool_uses_default(self):
+        assert make_plugin({"display_duration": True}).get_display_duration() == 15.0
+        assert make_plugin({"display_duration": False}).get_display_duration() == 15.0
+
+
+class TestValidateConfigDuration:
+    # validate_config must agree with get_display_duration about what a
+    # valid duration is — a config it accepts must not then be rejected
+    # (or silently defaulted) when the duration is actually read.
+
+    def test_positive_number_valid(self):
+        assert make_plugin({"display_duration": 20}).validate_config() is True
+
+    def test_zero_and_negative_invalid(self):
+        assert make_plugin({"display_duration": 0}).validate_config() is False
+        assert make_plugin({"display_duration": -5}).validate_config() is False
+
+    def test_bool_invalid(self):
+        # bool is an int subclass; True would otherwise pass as "positive
+        # number" here while get_display_duration rejects it.
+        assert make_plugin({"display_duration": True}).validate_config() is False
+        assert make_plugin({"display_duration": False}).validate_config() is False
+
+    def test_missing_duration_valid(self):
+        assert make_plugin({}).validate_config() is True

test/test_cache_cleanup_thread_ownership.py

@@ -0,0 +1,146 @@
+"""Tests that one cache directory gets one cleanup thread per process.
+
+The sweep lists a directory and deletes from it, so a second thread over the
+same directory only duplicates the scan. Nothing enforced that: every
+CacheManager started its own, and since the loop closes over `self`, a
+discarded manager could never be collected -- its thread stayed alive and
+re-scanned the same directory every 24 hours for the life of the process.
+
+On the dev rig a display process carried three, for one cache directory:
+
+    14:22:59.954  display_controller       (the real one)
+    14:22:59.973  startup validation, run 1  (discarded)
+    14:23:01.055  startup validation, run 2  (discarded)
+
+Startup validation runs twice and built a throwaway manager each time, purely
+to read a directory path.
+"""
+
+import threading
+
+import pytest
+
+from src.cache_manager import CacheManager
+
+
+@pytest.fixture(autouse=True)
+def _clean_registry():
+    CacheManager._cleanup_owners.clear()
+    yield
+    for owner in list(CacheManager._cleanup_owners.values()):
+        owner.stop_cleanup_thread()
+    CacheManager._cleanup_owners.clear()
+
+
+def _live_cleanup_threads():
+    return [t for t in threading.enumerate()
+            if t.name == 'DiskCacheCleanup' and t.is_alive()]
+
+
+@pytest.fixture
+def manager(tmp_path, monkeypatch):
+    """A CacheManager pinned to a temp dir, so tests never touch the real one."""
+    monkeypatch.setattr(CacheManager, '_get_writable_cache_dir',
+                        lambda self: str(tmp_path))
+    return CacheManager
+
+
+class TestOneThreadPerDirectory:
+    def test_a_single_manager_starts_one(self, manager):
+        before = len(_live_cleanup_threads())
+        m = manager()
+        try:
+            assert len(_live_cleanup_threads()) == before + 1
+        finally:
+            m.stop_cleanup_thread()
+
+    def test_three_managers_still_start_one(self, manager):
+        # Exactly the rig's shape: the real manager plus two throwaways.
+        before = len(_live_cleanup_threads())
+        managers = [manager() for _ in range(3)]
+        try:
+            assert len(_live_cleanup_threads()) == before + 1
+        finally:
+            for m in managers:
+                m.stop_cleanup_thread()
+
+    def test_the_first_one_owns_it(self, manager):
+        first, second = manager(), manager()
+        try:
+            assert CacheManager._cleanup_owners[first.cache_dir] is first
+            assert second._cleanup_thread is None
+        finally:
+            first.stop_cleanup_thread()
+            second.stop_cleanup_thread()
+
+    def test_the_survivor_can_take_over(self, manager):
+        first = manager()
+        first.stop_cleanup_thread()
+        assert not _live_cleanup_threads()
+
+        second = manager()
+        try:
+            # Ownership was released, so the directory is swept again rather
+            # than being left permanently unclaimed by a dead owner.
+            assert len(_live_cleanup_threads()) == 1
+            assert CacheManager._cleanup_owners[second.cache_dir] is second
+        finally:
+            second.stop_cleanup_thread()
+
+    def test_stopping_a_non_owner_does_not_unclaim_the_directory(self, manager):
+        first, second = manager(), manager()
+        try:
+            second.stop_cleanup_thread()      # never owned it
+            assert CacheManager._cleanup_owners[first.cache_dir] is first
+            assert len(_live_cleanup_threads()) == 1
+        finally:
+            first.stop_cleanup_thread()
+
+    def test_separate_directories_get_separate_threads(self, tmp_path, monkeypatch):
+        a, b = tmp_path / 'a', tmp_path / 'b'
+        a.mkdir()
+        b.mkdir()
+        dirs = iter([str(a), str(b)])
+        monkeypatch.setattr(CacheManager, '_get_writable_cache_dir',
+                            lambda self: next(dirs))
+        first, second = CacheManager(), CacheManager()
+        try:
+            assert first.cache_dir != second.cache_dir
+            assert len(_live_cleanup_threads()) == 2
+        finally:
+            first.stop_cleanup_thread()
+            second.stop_cleanup_thread()
+
+    def test_no_thread_leaks_across_many_constructions(self, manager):
+        before = len(_live_cleanup_threads())
+        made = [manager() for _ in range(12)]
+        try:
+            assert len(_live_cleanup_threads()) == before + 1
+        finally:
+            for m in made:
+                m.stop_cleanup_thread()
+        assert len(_live_cleanup_threads()) == before
+
+
+class TestValidatorDoesNotBuildItsOwn:
+    def test_it_uses_the_cache_manager_it_is_given(self, manager):
+        from src.startup_validator import StartupValidator
+
+        shared = manager()
+        try:
+            before = len(_live_cleanup_threads())
+            v = StartupValidator(config_manager=object(), cache_manager=shared)
+            v._validate_cache_directory()
+            assert len(_live_cleanup_threads()) == before, (
+                "validation started another cleanup thread")
+        finally:
+            shared.stop_cleanup_thread()
+
+    def test_without_one_it_cleans_up_after_itself(self, manager):
+        from src.startup_validator import StartupValidator
+
+        before = len(_live_cleanup_threads())
+        v = StartupValidator(config_manager=object())
+        v._validate_cache_directory()
+        assert len(_live_cleanup_threads()) == before, (
+            "the fallback manager left its cleanup thread running")

test/test_cache_manager.py

@@ -458,3 +458,26 @@ class TestDiskCacheWriteEconomy:
         cache = DiskCache(cache_dir=str(tmp_path))
         cache.set("k", {"when": datetime(2026, 7, 12, 10, 30)})
         assert cache.get("k") == {"when": "2026-07-12T10:30:00"}
+
+
+# --- the ceiling has to hold between cleanup sweeps ---------------------------
+
+def test_memory_cache_enforces_ceiling_on_every_write():
+    """_cleanup_memory_cache only runs every cleanup_interval seconds (300 by
+    default). If set() accepted entries without bound in between, a burst could
+    take the cache far past max_size -- which is the unbounded growth the limit
+    exists to prevent, and on a 1GB board the difference between a bounded cache
+    and a Pi that cannot fork.
+    """
+    from src.cache.memory_cache import MemoryCache
+
+    cache = MemoryCache(max_size=150, cleanup_interval=300.0)
+    for i in range(1000):
+        cache.set(f"k{i}", {"v": i})
+
+    assert len(cache._cache) <= 150
+    # The timestamp map has to be evicted alongside the values, or it becomes
+    # the leak instead.
+    assert len(cache._timestamps) <= 150
+    assert cache.get("k999") is not None, "the newest write must survive"
+    assert cache.get("k0") is None, "the oldest must be the one evicted"

test/test_cache_orphan_temp_sweep.py

@@ -0,0 +1,188 @@
+"""Tests that abandoned cache temp files get collected.
+
+DiskCache.set() writes through tempfile.mkstemp and os.replace, removing its
+own temp file in a finally. That covers a failed write, but not a process that
+dies between the two -- a SIGKILL, a lost restart race, a power cut, all
+ordinary on a Pi. Nothing collected what was left behind: the temp names are
+".<key>.json.<random>", and the expiry sweep only listed names ending in
+.json, so they accumulated for as long as the card had been in service.
+
+Measured on a live rig before this fix: 76 orphans totalling 1,050 MB -- 81%
+of the entire cache directory -- the oldest six months old.
+
+The predicate that decides what to delete is tested harder than the sweep
+itself, because a false positive here destroys real data.
+"""
+
+import os
+import time
+
+import pytest
+
+from src.cache.disk_cache import DiskCache, _ORPHAN_TEMP_MAX_AGE_SECONDS
+
+
+class FakeStrategy:
+    @staticmethod
+    def get_data_type_from_key(key):
+        return 'default'
+
+
+POLICIES = {'default': 30}
+
+
+@pytest.fixture
+def cache(tmp_path):
+    return DiskCache(str(tmp_path))
+
+
+def _age(path, seconds):
+    old = time.time() - seconds
+    os.utime(path, (old, old))
+
+
+def _write(tmp_path, name, body='{}'):
+    p = tmp_path / name
+    p.write_text(body, encoding='utf-8')
+    return p
+
+
+class TestWhatCountsAsAnOrphan:
+    @pytest.mark.parametrize('name', [
+        '.weather.json.a1b2c3d4',
+        '.odds_espn_football_nfl_401.json.xyz00000',
+        '.a.json.b',
+    ])
+    def test_our_temp_files_are_orphans(self, name):
+        assert DiskCache._is_orphaned_temp(name)
+
+    @pytest.mark.parametrize('name', [
+        'weather.json',            # real data
+        '.weather.json',           # a dotted key that completed
+        '.gitignore',              # not ours
+        '.hidden',                 # not ours
+        'weather.json.bak',        # no leading dot: someone else's
+        '.json.abc',               # no key between the dot and .json.
+        '.weather.json.',          # no random component
+        'notes.txt',
+    ])
+    def test_everything_else_is_left_alone(self, name):
+        assert not DiskCache._is_orphaned_temp(name)
+
+    def test_the_names_set_actually_creates_are_matched(self, cache, tmp_path):
+        """Guard against the predicate and the writer drifting apart."""
+        created = []
+        real = os.replace
+
+        def capture(src, dst):
+            created.append(os.path.basename(src))
+            return real(src, dst)
+
+        import src.cache.disk_cache as mod
+        mod.os.replace = capture
+        try:
+            cache.set('weather', {'v': 1})
+        finally:
+            mod.os.replace = real
+
+        assert created, "set() did not go through the temp-file path"
+        assert all(DiskCache._is_orphaned_temp(n) for n in created), created
+
+
+class TestTheSweep:
+    def test_an_old_orphan_is_removed(self, cache, tmp_path):
+        p = _write(tmp_path, '.weather.json.a1b2c3d4', 'x' * 5000)
+        _age(p, _ORPHAN_TEMP_MAX_AGE_SECONDS + 60)
+
+        stats = cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+
+        assert not p.exists()
+        assert stats['orphan_temp_files_deleted'] == 1
+        assert stats['space_freed_bytes'] >= 5000
+
+    def test_an_in_flight_write_is_not_snatched_away(self, cache, tmp_path):
+        # The whole risk of this sweep: deleting a temp file another thread is
+        # about to os.replace into place.
+        p = _write(tmp_path, '.weather.json.inflight')
+
+        cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+
+        assert p.exists()
+
+    def test_real_cache_files_survive(self, cache, tmp_path):
+        fresh = _write(tmp_path, 'weather.json')
+        dotted = _write(tmp_path, '.weather.json')
+        _age(dotted, _ORPHAN_TEMP_MAX_AGE_SECONDS + 60)
+
+        cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+
+        assert fresh.exists()
+        assert dotted.exists(), "a completed .json was treated as a temp file"
+
+    def test_unrelated_dotfiles_survive(self, cache, tmp_path):
+        keep = _write(tmp_path, '.gitignore')
+        _age(keep, 400 * 86400)
+
+        cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+
+        assert keep.exists()
+
+    def test_expiry_still_works_alongside_it(self, cache, tmp_path):
+        stale = _write(tmp_path, 'old.json')
+        _age(stale, 40 * 86400)          # past the 30-day default
+        orphan = _write(tmp_path, '.old.json.zz999999')
+        _age(orphan, _ORPHAN_TEMP_MAX_AGE_SECONDS + 60)
+
+        stats = cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+
+        assert not stale.exists()
+        assert not orphan.exists()
+        assert stats['files_deleted'] == 2
+        assert stats['orphan_temp_files_deleted'] == 1
+
+    def test_the_rig_scenario(self, cache, tmp_path):
+        """76 orphans of assorted ages, none of them reachable before."""
+        for i in range(76):
+            p = _write(tmp_path, '.sched_%d.json.r%06d' % (i, i), 'x' * 1000)
+            _age(p, (i + 2) * 86400)
+        keep = _write(tmp_path, 'sched.json')
+
+        stats = cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+
+        assert stats['orphan_temp_files_deleted'] == 76
+        assert keep.exists()
+        assert not list(tmp_path.glob('.sched_*'))
+        # The summary line is "<deleted>/<scanned>", so an orphan that is
+        # deleted but never counted as scanned renders as "76/1".
+        assert stats['files_scanned'] == 77
+        assert stats['files_deleted'] <= stats['files_scanned']
+
+    def test_deleted_never_exceeds_scanned(self, cache, tmp_path):
+        p = _write(tmp_path, '.only.json.a1b2c3d4')
+        _age(p, _ORPHAN_TEMP_MAX_AGE_SECONDS + 60)
+
+        stats = cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+
+        assert stats['files_deleted'] == 1
+        assert stats['files_scanned'] == 1
+
+    def test_a_missing_file_mid_sweep_is_not_an_error(self, cache, tmp_path):
+        p = _write(tmp_path, '.weather.json.a1b2c3d4')
+        _age(p, _ORPHAN_TEMP_MAX_AGE_SECONDS + 60)
+
+        import src.cache.disk_cache as mod
+        real = mod.os.path.getsize
+
+        def vanish(path):
+            if path.endswith('.a1b2c3d4'):
+                os.remove(path)
+                raise FileNotFoundError(path)
+            return real(path)
+
+        mod.os.path.getsize = vanish
+        try:
+            stats = cache.cleanup_expired_files(FakeStrategy(), POLICIES)
+        finally:
+            mod.os.path.getsize = real
+
+        assert stats['errors'] == 0

test/test_cache_ttl_honoured.py

@@ -0,0 +1,130 @@
+"""Tests that a per-entry ttl actually controls expiry.
+
+Regression under test: `CacheManager.set(key, data, ttl=...)` stored the value
+and no read path ever consulted it. Expiry came from a `max_age` inferred from
+substrings in the key ("live", "odds", "stock"), so every caller passing `ttl=`
+-- 48 sites across the plugins and 4 in the core -- was writing a number that
+did nothing. The old docstring admitted as much: "stored for compatibility but
+expiration is still controlled via max_age when reading".
+
+Measured against a real device's cache (8,873 entries carrying a ttl), the
+inferred value and the intended one disagreed almost everywhere:
+
+    stocks    max_age  600  vs ttl   1800   4903 entries
+    news      max_age 3600  vs ttl    600   1770 entries
+    odds      max_age 1800  vs ttl   3600   1301 entries
+    images    max_age  300  vs ttl 2592000    20 entries
+
+No `sports_live` entry carries a ttl, so live scores keep their inferred
+30-second freshness either way.
+"""
+
+import time
+
+import pytest
+
+from src.cache.memory_cache import MemoryCache
+from src.cache.disk_cache import DiskCache
+
+
+@pytest.fixture
+def disk(tmp_path):
+    return DiskCache(cache_dir=str(tmp_path))
+
+
+def _record(ttl=None, age=0.0):
+    rec = {"data": {"v": 1}, "timestamp": time.time() - age}
+    if ttl is not None:
+        rec["ttl"] = ttl
+    return rec
+
+
+class TestDiskCacheHonoursTtl:
+    def test_ttl_longer_than_max_age_keeps_the_entry(self, disk):
+        # The odds case: written wanting an hour, expired at 30 minutes.
+        disk.set("odds_espn_football_nfl_401", _record(ttl=3600, age=1900))
+        assert disk.get("odds_espn_football_nfl_401", max_age=1800) is not None
+
+    def test_ttl_shorter_than_max_age_expires_the_entry(self, disk):
+        # The news case: written wanting 10 minutes, kept for an hour.
+        disk.set("news_NHL_1", _record(ttl=600, age=900))
+        assert disk.get("news_NHL_1", max_age=3600) is None
+
+    def test_without_a_ttl_max_age_still_applies(self, disk):
+        disk.set("plain_key", _record(age=400))
+        assert disk.get("plain_key", max_age=300) is None
+        disk.set("plain_key2", _record(age=100))
+        assert disk.get("plain_key2", max_age=300) is not None
+
+    def test_a_fresh_entry_within_its_ttl_survives(self, disk):
+        disk.set("k", _record(ttl=600, age=10))
+        assert disk.get("k", max_age=30) is not None
+
+    def test_ttl_zero_expires_immediately(self, disk):
+        # 0 means zero seconds, not "forever" -- max_age=None is how a caller
+        # asks for no expiry.
+        disk.set("k", _record(ttl=0, age=1))
+        assert disk.get("k", max_age=99999) is None
+
+    @pytest.mark.parametrize("bad", ["600", None, True, False, -5, {"a": 1}])
+    def test_a_nonsense_ttl_falls_back_to_max_age(self, disk, bad):
+        # Including bools: True is an int in Python and must not become a 1s ttl.
+        rec = _record(age=400)
+        rec["ttl"] = bad
+        disk.set("k_%s" % type(bad).__name__, rec)
+        assert disk.get("k_%s" % type(bad).__name__, max_age=300) is None
+
+
+class TestMemoryCacheHonoursTtl:
+    def test_ttl_longer_than_max_age_keeps_the_entry(self):
+        m = MemoryCache()
+        m.set("k", _record(ttl=3600))
+        m._timestamps["k"] = time.time() - 1900
+        assert m.get("k", max_age=1800) is not None
+
+    def test_ttl_shorter_than_max_age_expires_the_entry(self):
+        m = MemoryCache()
+        m.set("k", _record(ttl=600))
+        m._timestamps["k"] = time.time() - 900
+        assert m.get("k", max_age=3600) is None
+
+    def test_without_a_ttl_max_age_still_applies(self):
+        m = MemoryCache()
+        m.set("k", _record())
+        m._timestamps["k"] = time.time() - 400
+        assert m.get("k", max_age=300) is None
+
+    def test_both_layers_agree(self, tmp_path):
+        """A record must not be live in one layer and expired in the other."""
+        rec = _record(ttl=3600, age=1900)
+        d = DiskCache(cache_dir=str(tmp_path))
+        d.set("k", rec)
+        m = MemoryCache()
+        m.set("k", rec)
+        m._timestamps["k"] = rec["timestamp"]
+        assert (d.get("k", max_age=1800) is not None) == (m.get("k", max_age=1800) is not None)
+
+
+class TestEndToEnd:
+    def test_set_then_get_respects_the_ttl(self, tmp_path, monkeypatch):
+        """The behaviour a caller of CacheManager.set(ttl=...) expects."""
+        from src.cache_manager import CacheManager
+
+        cm = CacheManager()
+        cm._disk_cache_component = DiskCache(cache_dir=str(tmp_path))
+        cm._memory_cache_component = MemoryCache()
+
+        cm.set("odds_espn_football_nfl_401", {"spread": 6.5}, ttl=3600)
+
+        # Age the stored record past the inferred max_age for odds (1800s) but
+        # within the ttl the caller asked for.
+        path = cm._disk_cache_component.get_cache_path("odds_espn_football_nfl_401")
+        import json
+        rec = json.load(open(path))
+        rec["timestamp"] = time.time() - 1900
+        json.dump(rec, open(path, "w"))
+        cm._memory_cache_component.clear() if hasattr(
+            cm._memory_cache_component, "clear") else None
+
+        got = cm.get_with_auto_strategy("odds_espn_football_nfl_401")
+        assert got is not None, "the ttl the caller asked for was ignored"

test/test_checkbox_group_stale_values.py

@@ -0,0 +1,113 @@
+"""A checkbox group must not post back options it cannot show.
+
+The enum that lets the widget draw checkboxes is also what validates the
+saved value. When a league retires a team code -- OAK for the Athletics, ARI
+for the Coyotes -- or a schema drops an option, a config that still holds the
+old value has nothing to render for it. The value stayed in the hidden
+``_data`` input regardless, because that input is seeded from the stored array
+and only rebuilt by ``updateCheckboxGroupData()`` on change. Editing any other
+field on that plugin therefore posted the stale value back, the schema
+rejected it, and the save endpoint returned 400
+``CONFIG_VALIDATION_FAILED`` -- so the whole plugin became uneditable until
+the user worked out which invisible entry was at fault.
+
+Runtime was never affected: plugin loading treats schema violations as
+warn/degrade, and the stale code already matched no team. Only the web UI
+blocked.
+
+These tests render the checkbox-group block lifted *out of the shipped
+template*, following test_enum_option_labels.py, so they exercise the
+production expression rather than a copy that could drift from it.
+"""
+import json
+import re
+from pathlib import Path
+
+from jinja2 import DictLoader, Environment
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+CONFIG_FORM = (PROJECT_ROOT / 'web_interface' / 'templates' / 'v3' / 'partials'
+               / 'plugin_config.html')
+
+# The checkbox-group branch: from its `{% elif %}` guard through the sentinel
+# hidden input that closes it. Anchored on the guard so the match cannot run on
+# into a neighbouring widget branch.
+BLOCK_RE = re.compile(
+    r"\{%\s*elif x_widget == 'checkbox-group'\s*%\}(.*?)"
+    r"<input type=\"hidden\" name=\"\{\{ full_key \}\}\[\]\" value=\"\">",
+    re.S,
+)
+
+
+def _shipped_block() -> str:
+    """Return the live checkbox-group block lifted from plugin_config.html."""
+    source = CONFIG_FORM.read_text(encoding='utf-8')
+    match = BLOCK_RE.search(source)
+    assert match, (
+        'could not find the checkbox-group block in plugin_config.html — the '
+        'template changed shape and this guard needs updating'
+    )
+    block = match.group(1)
+    assert 'data-option-value' in block, 'extracted the wrong branch'
+    assert '{% elif' not in block, 'extraction ran past the checkbox-group branch'
+    return block
+
+
+def _render(prop: dict, value=None) -> str:
+    env = Environment(loader=DictLoader({'f': _shipped_block()}), autoescape=True)
+    return env.get_template('f').render(
+        prop=prop, value=value, field_id='fid', full_key='k'
+    )
+
+
+def _submitted(html: str) -> list:
+    """The array the form will actually post: the hidden _data input."""
+    match = re.search(r'id="fid_data"[^>]*\svalue=\'([^\']*)\'', html)
+    assert match, f'hidden _data input not found in:\n{html}'
+    return json.loads(match.group(1).replace('&#39;', "'"))
+
+
+def _checked(html: str) -> list:
+    return re.findall(r'data-option-value="([^"]+)"[^>]*checked', html)
+
+
+MLB = {'type': 'array', 'items': {'type': 'string', 'enum': ['NYY', 'BOS', 'ATH']},
+       'x-widget': 'checkbox-group'}
+
+
+def test_a_retired_code_is_not_posted_back() -> None:
+    """The regression: OAK became ATH, and OAK used to ride along on save."""
+    html = _render(MLB, ['NYY', 'OAK'])
+    assert _submitted(html) == ['NYY'], 'stale value would still be submitted'
+
+
+def test_the_dropped_value_is_named_rather_than_vanishing() -> None:
+    html = _render(MLB, ['NYY', 'OAK'])
+    assert 'OAK' in html
+    assert 'data-stale-options' in html
+
+
+def test_valid_values_are_untouched_and_still_checked() -> None:
+    html = _render(MLB, ['NYY', 'ATH'])
+    assert _submitted(html) == ['NYY', 'ATH']
+    assert sorted(_checked(html)) == ['ATH', 'NYY']
+    assert 'data-stale-options' not in html
+
+
+def test_an_all_stale_selection_clears_rather_than_blocking() -> None:
+    html = _render(MLB, ['OAK', 'SD'])
+    assert _submitted(html) == []
+
+
+def test_an_empty_enum_leaves_the_value_alone() -> None:
+    """No options means nothing to validate against — filtering would wipe it."""
+    prop = {'type': 'array', 'items': {'type': 'string'}, 'x-widget': 'checkbox-group'}
+    html = _render(prop, ['ANYTHING', 'GOES'])
+    assert _submitted(html) == ['ANYTHING', 'GOES']
+
+
+def test_unset_value_falls_back_to_the_default() -> None:
+    prop = dict(MLB, default=['BOS'])
+    html = _render(prop, None)
+    assert _submitted(html) == ['BOS']
+    assert _checked(html) == ['BOS']

test/test_composer_blueprint_registered.py

@@ -0,0 +1,64 @@
+"""composer_bp must actually be registered on the Flask app.
+
+web_interface/blueprints/composer.py defines a full blueprint -- the
+/composer/ page, /composer/api/generate, /composer/api/install, etc. -- but
+nothing in web_interface/app.py imported or registered it. pages_v3 and
+api_v3 both go through the same "import, wire managers, register_blueprint"
+sequence; composer_bp had none of the three. The routes existed and worked
+in isolation (every other composer test builds its own minimal Flask app
+and registers composer_bp directly), but in the real running app every
+composer URL -- including the page itself -- was a 404, and the frontend's
+hardcoded '/composer/api/...' fetches (composer-app.js) had nothing to
+reach.
+
+app.py has import-time side effects (it loads config, builds managers, and
+in some configurations starts background threads), so this scans the source
+text rather than importing the module -- the same approach already used for
+test_composer_js_contracts.py's structural checks.
+"""
+import re
+from pathlib import Path
+
+APP_PY = Path(__file__).resolve().parent.parent / "web_interface" / "app.py"
+
+
+def _source() -> str:
+    return APP_PY.read_text()
+
+
+def test_composer_blueprint_is_imported():
+    src = _source()
+    assert re.search(
+        r"from web_interface\.blueprints\.composer import composer_bp", src
+    ), "app.py never imports composer_bp"
+
+
+def test_composer_blueprint_is_registered_under_composer_prefix():
+    src = _source()
+    match = re.search(
+        r"app\.register_blueprint\(\s*composer_bp\s*,\s*url_prefix\s*=\s*"
+        r"['\"](?P<prefix>[^'\"]*)['\"]",
+        src,
+    )
+    assert match, "app.py never calls app.register_blueprint(composer_bp, ...)"
+    # composer.html and composer-app.js hardcode '/composer/api/...' and
+    # '/composer/api/fonts/...' -- any other prefix serves a page whose own
+    # fetches and @font-face all 404.
+    assert match.group("prefix") == "/composer"
+
+
+def test_composer_blueprint_managers_are_wired_before_registration():
+    """Registered with no managers set means every route 503s or crashes.
+
+    composer.py initialises config_manager/plugin_manager/plugins_dir/
+    project_root to None at import time specifically so app.py can set them
+    -- the same pattern api_v3 and pages_v3 use.
+    """
+    src = _source()
+    register_match = re.search(r"app\.register_blueprint\(\s*composer_bp", src)
+    assert register_match, "app.py never registers composer_bp"
+    before_registration = src[: register_match.start()]
+    for attr in ("config_manager", "plugin_manager", "plugins_dir", "project_root"):
+        assert re.search(rf"composer_bp\.{attr}\s*=", before_registration), (
+            f"composer_bp.{attr} is never set before registration"
+        )

test/test_composer_code_injection.py

@@ -0,0 +1,323 @@
+"""The composer generates Python that the plugin loader imports and executes.
+
+/api/install writes the generated manager.py into plugins_dir and the loader
+imports it, so anything the payload can splice into that source runs on the
+device. The ast.parse check in _generate_plugin_files rejects only *invalid*
+syntax -- an injected `import os` is perfectly valid and passed it.
+
+Two ways in, both confirmed against the code before it was fixed:
+
+  metadata.name = a name containing a triple-quote, a newline, then
+                  `import os; PWNED = os.getuid()`, then another triple-quote
+      -> closes the module docstring; the rest became module-level statements
+         (spelled out rather than shown literally -- writing the payload into
+         this docstring closes *this* file's docstring, which is the bug)
+
+  element x     = '0 or __import__("os").system("id")'
+      -> f-string interpolated it verbatim: x=0 or __import__("os").system("id")
+"""
+import ast
+import re
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from web_interface.blueprints import composer as C  # noqa: E402
+
+BASE_META = {"id": "test-plugin", "name": "Clock", "author": "a",
+             "version": "1.0.0", "description": "d"}
+
+#: Values that terminate a Python expression and start a new statement.
+EXPR_PAYLOADS = [
+    '0 or __import__("os").system("id")',
+    '0);import os;os.system("id");(',
+    '__import__("subprocess").run(["id"])',
+    "0 if False else exec('x=1')",
+    "1e999", "nan", "0x41", "0__0",
+]
+
+#: Values that close a string literal in the generated source.
+LITERAL_PAYLOADS = [
+    'Clock"""\nimport os; PWNED = os.getuid()\n"""',
+    "Clock'''\nimport os\n'''",
+    'Clock" + __import__("os").system("id") + "',
+    "Clock\\", "Clock\nimport os",
+]
+
+
+def _payload(**over):
+    # dataModel.configVars is the key _generate_plugin_files reads; "config_vars"
+    # was never looked at, so anything passed through it tested nothing.
+    p = {"metadata": dict(BASE_META), "elements": [],
+         "dataModel": {"configVars": over.pop("config_vars", [])}}
+    p["metadata"].update(over.pop("metadata", {}))
+    p.update(over)
+    return p
+
+
+def _generated(payload):
+    return C._generate_plugin_files(payload)["manager.py"]
+
+
+def _module_level_code(src):
+    """Statements at module level that are not the docstring/imports/classes."""
+    tree = ast.parse(src)
+    out = []
+    for node in tree.body:
+        if isinstance(node, (ast.ClassDef, ast.FunctionDef, ast.ImportFrom, ast.Import)):
+            continue
+        if isinstance(node, ast.Expr) and isinstance(node.value, ast.Constant):
+            continue  # the docstring
+        out.append(ast.unparse(node))
+    return out
+
+
+@pytest.mark.parametrize("payload", LITERAL_PAYLOADS)
+def test_a_name_that_breaks_out_of_a_literal_is_refused(payload):
+    with pytest.raises(C.ComposerInputError):
+        _generated(_payload(metadata={"name": payload}))
+
+
+#: Types with a drawing branch in manager.py.j2. An injection test using any
+#: other type proves nothing: _preprocess_elements drops it, so its values
+#: never reach the generated source and every assertion passes trivially.
+#: This test previously used "line", which has never had a branch.
+RENDERED_GEOMETRY_CASES = [
+    ("rectangle", {"x": 0, "y": 0, "width": 10, "height": 8}),
+    ("arc", {"x": 0, "y": 0, "width": 24, "height": 24}),
+    ("ellipse", {"x": 0, "y": 0, "width": 24, "height": 12}),
+    ("rounded_rectangle", {"x": 0, "y": 0, "width": 24, "height": 10}),
+    ("gauge", {"x": 0, "y": 0, "width": 32, "height": 32}),
+]
+
+
+@pytest.mark.parametrize("etype,base", RENDERED_GEOMETRY_CASES)
+@pytest.mark.parametrize("evil", EXPR_PAYLOADS)
+@pytest.mark.parametrize("field", ["x", "y", "width", "height"])
+def test_a_non_numeric_geometry_value_cannot_reach_the_source(etype, base, evil, field):
+    """width/height were interpolated raw into the generated source.
+
+    p['x2_expr'] = f"({x_expr}) + {w}" with w straight off the payload, so a
+    rectangle with width='0 or __import__("os").system("id")' produced
+
+        [0, 0, (0) + 0 or __import__("os").system("id"), (0) + 8],
+
+    in a manager.py that /api/install writes to disk and the loader imports.
+    """
+    el = {"type": etype, "id": "e1", **base}
+    el[field] = evil
+    src = _generated(_payload(elements=[el]))
+    assert "__import__" not in src, f"{etype}.{field}={evil!r} reached the generated source"
+    assert "os.system" not in src
+    assert not _module_level_code(src), \
+        f"{etype}.{field}={evil!r} produced module-level statements: {_module_level_code(src)}"
+
+
+def test_every_injection_case_uses_a_type_that_actually_renders():
+    """Guards against the whole suite quietly going vacuous again.
+
+    An element type with no template branch is dropped before generation, so
+    an injection test written against one asserts nothing and still passes.
+    """
+    used = {etype for etype, _ in RENDERED_GEOMETRY_CASES}
+    missing = used - set(C._RENDERABLE_ELEMENT_TYPES)
+    assert not missing, f"injection tests use non-rendering types: {sorted(missing)}"
+
+
+@pytest.mark.parametrize("evil", EXPR_PAYLOADS)
+@pytest.mark.parametrize("channel", ["r", "g", "b"])
+def test_a_non_numeric_colour_channel_cannot_reach_the_source(evil, channel):
+    el = {"type": "text", "id": "t1", "x": 0, "y": 0, "text": "hi",
+          "font": "press_start", "r": 255, "g": 255, "b": 255}
+    el[channel] = evil
+    src = _generated(_payload(elements=[el]))
+    assert "__import__" not in src and "os.system" not in src
+    assert not _module_level_code(src)
+
+
+def test_colour_channels_are_clamped_to_a_byte():
+    el = {"type": "text", "id": "t1", "x": 0, "y": 0, "text": "hi",
+          "font": "press_start", "r": 99999, "g": -5, "b": 128}
+    src = _generated(_payload(elements=[el]))
+    assert "(255, 0, 128)" in src, "channels were not clamped to 0-255"
+
+
+#: fillR/fillG/fillB (and outR/G/B, bgR/G/B, trackR/G/B) go through
+#: _as_fill_filter -> _as_rgb_filter, a separate path from _rgb_expr above.
+#: It used raw int() until it was found to raise ValueError on a non-numeric
+#: channel instead of clamping like every other coerced value in this module.
+@pytest.mark.parametrize("evil", EXPR_PAYLOADS)
+@pytest.mark.parametrize("channel", ["fillR", "fillG", "fillB"])
+def test_a_non_numeric_fill_channel_cannot_reach_the_source(evil, channel):
+    el = {"type": "rectangle", "id": "r1", "x": 0, "y": 0, "width": 10, "height": 8,
+          "fillR": 0, "fillG": 0, "fillB": 128}
+    el[channel] = evil
+    src = _generated(_payload(elements=[el]))
+    assert "__import__" not in src and "os.system" not in src
+    assert not _module_level_code(src)
+
+
+def test_fill_channels_are_clamped_to_a_byte():
+    el = {"type": "rectangle", "id": "r1", "x": 0, "y": 0, "width": 10, "height": 8,
+          "fillR": 99999, "fillG": -5, "fillB": 128}
+    src = _generated(_payload(elements=[el]))
+    assert "(255, 0, 128)" in src, "fill channels were not clamped to 0-255"
+
+
+def test_the_generated_module_still_has_no_top_level_statements():
+    """The clean case: a normal payload produces only imports and a class."""
+    el = {"type": "text", "id": "t1", "x": 4, "y": 4, "text": "hi",
+          "font": "press_start", "r": 1, "g": 2, "b": 3}
+    src = _generated(_payload(elements=[el]))
+    assert not _module_level_code(src)
+    assert "(1, 2, 3)" in src
+
+
+# --- config variable keys ---------------------------------------------------
+
+def _with_key(key):
+    return {"metadata": dict(BASE_META), "elements": [],
+            "dataModel": {"configVars": [{"key": key, "type": "string",
+                                          "default": "x", "label": "L"}]}}
+
+
+@pytest.mark.parametrize("key", ["class", "def", "import", "None", "True",
+                                 "lambda", "pass", "match", "case"])
+def test_a_keyword_config_key_is_named_in_the_error(key):
+    """ast.parse already rejected these, but as an unhelpful line number.
+
+    "Generated code has a syntax error: invalid syntax (line 17)" tells the
+    user nothing about which field to fix.
+    """
+    with pytest.raises(C.ComposerInputError) as exc:
+        _generated(_with_key(key))
+    assert key in str(exc.value) and "keyword" in str(exc.value).lower()
+
+
+@pytest.mark.parametrize("key", ["config", "logger", "display_manager",
+                                 "cache_manager", "plugin_id", "enabled",
+                                 "self", "update", "display"])
+def test_a_reserved_attribute_config_key_is_refused(key):
+    """These generate *valid* Python that silently clobbers plugin state.
+
+    The worst is `config`: the assignment lands right after super().__init__(),
+    so `self.config = config.get("config", "x")` replaces the plugin's config
+    dict with a string and every later self.config.get(...) fails at runtime.
+    """
+    with pytest.raises(C.ComposerInputError) as exc:
+        _generated(_with_key(key))
+    assert key in str(exc.value) and "reserved" in str(exc.value).lower()
+
+
+@pytest.mark.parametrize("key", ["brightness", "my_var", "_private", "x1",
+                                 "update_interval_seconds"])
+def test_ordinary_config_keys_are_still_accepted(key):
+    src = _generated(_with_key(key))
+    assert f"self.{key} = config.get(" in src
+
+
+def test_the_generated_config_assignment_does_not_precede_super_init():
+    """Guards the reasoning behind the reserved list, not just the list."""
+    src = _generated(_with_key("brightness"))
+    body = src.splitlines()
+    super_at = next(i for i, line in enumerate(body) if "super().__init__(" in line)
+    assign_at = next(i for i, line in enumerate(body)
+                     if "self.brightness = config.get(" in line)
+    assert assign_at > super_at, (
+        "config vars are assigned before super().__init__(); the reserved-name "
+        "list assumes they land after it")
+
+
+# --- optional keys ----------------------------------------------------------
+
+@pytest.mark.parametrize("el_type,missing", [
+    ("text", "text"), ("text", "text2"), ("clock", "format"),
+])
+def test_an_element_missing_an_optional_key_does_not_500(el_type, missing):
+    """`p` is a copy of the raw element, so an absent key stays absent.
+
+    The defaults were applied to locals only, so manager.py.j2 rendered
+    `{{ el.text | tojson }}` over a jinja2.Undefined and tojson raised
+    TypeError -- which no handler catches, making a missing key a 500 rather
+    than a validation error or a sensible default.
+    """
+    el = {"type": el_type, "id": "e1", "x": 0, "y": 0, "font": "press_start"}
+    src = _generated(_payload(elements=[el]))
+    ast.parse(src)          # must still be valid Python
+    assert "Undefined" not in src
+
+
+def test_a_clock_without_a_format_uses_the_documented_default():
+    el = {"type": "clock", "id": "c1", "x": 0, "y": 0, "font": "press_start"}
+    src = _generated(_payload(elements=[el]))
+    assert '"%H:%M"' in src, "the %H:%M default did not reach the generated source"
+
+
+#: (element type, channel key, base element) for colour channels that were
+#: interpolated raw rather than through _rgb_expr/_safe_int. Prefixed channels
+#: (emptyR/G/B, labelR/G/B) were the ones the original r/g/b test never reached.
+RAW_COLOUR_CASES = [
+    ("progress_bar", "r",      {"x": 0, "y": 0}),
+    ("progress_bar", "g",      {"x": 0, "y": 0}),
+    ("pips",         "b",      {"x": 0, "y": 0}),
+    ("pips",         "emptyR", {"x": 0, "y": 0}),
+    ("pips",         "emptyG", {"x": 0, "y": 0}),
+    ("sparkline",    "r",      {"x": 0, "y": 0}),
+    ("gauge",        "labelR", {"x": 0, "y": 0, "width": 32, "height": 32}),
+    ("gauge",        "labelB", {"x": 0, "y": 0, "width": 32, "height": 32}),
+]
+
+
+@pytest.mark.parametrize("etype,channel,base", RAW_COLOUR_CASES)
+@pytest.mark.parametrize("evil", EXPR_PAYLOADS)
+def test_a_prefixed_colour_channel_cannot_reach_the_source(etype, channel, base, evil):
+    """Five tuples were built with f"({el.get('r', 100)}, ...)" -- no coercion.
+
+    The pre-existing colour test only covered r/g/b on a text element, so the
+    prefixed channels and the four other types were never exercised.
+    """
+    el = {"type": etype, "id": "e1", **base}
+    el[channel] = evil
+    src = _generated(_payload(elements=[el]))
+    assert "__import__" not in src, f"{etype}.{channel}={evil!r} reached the source"
+    assert "os.system" not in src
+    assert not _module_level_code(src)
+
+
+@pytest.mark.parametrize("value", [float("inf"), float("-inf"), float("nan")])
+@pytest.mark.parametrize("field", ["x", "y", "width", "height"])
+def test_a_non_finite_dimension_does_not_escape_as_an_unhandled_error(field, value):
+    """json.loads accepts Infinity/NaN and Flask passes them through, so a
+    payload can hand _safe_int a non-finite float. int(inf) raises
+    OverflowError -- neither ValueError nor ComposerInputError -- so it escaped
+    both handlers and surfaced as a 500 with a traceback instead of a 422."""
+    el = {"type": "rectangle", "id": "r1", "x": 0, "y": 0, "width": 10, "height": 8}
+    el[field] = value
+    src = _generated(_payload(elements=[el]))       # must not raise
+    # A non-finite value must be replaced by the default, not spelled into the
+    # source. Word-boundary match: "info" in self.logger.info contains "inf".
+    assert not re.search(r"\b(inf|nan|Infinity|NaN)\b", src), \
+        f"{field}={value!r} leaked a non-finite literal into the source"
+    assert not _module_level_code(src)
+
+
+@pytest.mark.parametrize("bad_id", [
+    'x = __import__("os").system("id") #',
+    "x\nimport os\n_y",
+    "x[0]",
+    "",
+    "a" * 200,
+])
+def test_a_marquee_id_cannot_become_code(bad_id):
+    """data_key is spliced UNQUOTED into variable names
+    (_{{ data_key }}_text = ...), so a non-identifier id landed in the source
+    as code. ast.parse caught it, but the caller then got an opaque
+    "Generated code has a syntax error" rather than being told the id is bad."""
+    el = {"type": "marquee", "id": bad_id, "x": 0, "y": 0, "text": "hi"}
+    src = _generated(_payload(elements=[el]))       # must not raise
+    assert "__import__(" not in src
+    assert "os.system(" not in src
+    assert not _module_level_code(src)

test/test_composer_empty_block.py

@@ -0,0 +1,124 @@
+"""An element the template cannot draw must not produce an empty `if` block.
+
+manager.py.j2 wraps each element in `if width >= N:` (breakpoint) and/or
+`if int(time.time() * 2) % 2:` (blink), and the body comes from the per-type
+branches. A type with no branch contributed nothing, so the wrapper opened a
+block with no statements in it. ast.parse in _generate_plugin_files then
+failed and the caller was told only:
+
+    Generated code has a syntax error: expected an indented block after
+    'if' statement on line 49
+
+which names a line of generated source the user never sees. Confirmed against
+the code before the fix with a `group` element carrying minWidth.
+
+Two defences, both covered here: _preprocess_elements drops types the template
+has no branch for, and the template emits a `pass` fallback so a type added to
+the canvas before its branch exists degrades to a no-op instead of a broken
+plugin.
+"""
+import ast
+import re
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from web_interface.blueprints import composer as C  # noqa: E402
+
+TEMPLATE = (Path(__file__).resolve().parent.parent
+            / "web_interface/templates/v3/composer/manager.py.j2")
+
+BASE_META = {"id": "test-plugin", "name": "Clock", "author": "a",
+             "version": "1.0.0", "description": "d"}
+
+
+def generate(element):
+    return C._generate_plugin_files({
+        "metadata": BASE_META,
+        "elements": [element],
+        "dataModel": {"configVars": []},
+    })
+
+
+@pytest.mark.parametrize("wrapper", [
+    {"minWidth": 64},          # breakpoint block
+    {"blink": True},           # blink block
+    {"minWidth": 64, "blink": True},   # both, nested
+])
+@pytest.mark.parametrize("etype", ["group", "widget_9000", "section"])
+def test_undrawable_element_does_not_break_generation(etype, wrapper):
+    element = {"type": etype, "x": 0, "y": 0, "color": "#ffffff", **wrapper}
+    files = generate(element)          # must not raise ComposerInputError
+    assert "manager.py" in files
+
+
+def test_drawable_element_still_renders_inside_a_breakpoint():
+    files = generate({"type": "text", "text": "hi", "x": 0, "y": 0,
+                      "minWidth": 64, "color": "#ffffff"})
+    src = files["manager.py"]
+    assert "if width >= 64:" in src
+    assert "draw_text" in src
+
+
+def test_renderable_types_match_the_template_branches():
+    """The constant and the template must agree.
+
+    A type listed in the constant with no branch emits an empty block (the bug
+    above); a type with a branch but missing from the constant is silently
+    dropped from every generated plugin. Neither is visible without this check.
+    """
+    branches = set(re.findall(r"el\.type == '([a-z_]+)'", TEMPLATE.read_text()))
+    assert branches == set(C._RENDERABLE_ELEMENT_TYPES)
+
+
+def test_template_closes_the_branch_chain_with_a_fallback():
+    """Belt and braces: even if the constant drifts, no empty block escapes."""
+    text = TEMPLATE.read_text()
+    assert "{% else %}" in text
+    assert "pass  # element type" in text
+
+
+@pytest.mark.parametrize("wrapper", [
+    {"minWidth": 64},
+    {"blink": True},
+    {"minWidth": 64, "blink": True},
+])
+def test_dynamic_text_with_non_config_binding_does_not_break_generation(wrapper):
+    """dynamic_text only renders a body for binding.source == 'config'.
+
+    _preprocess_elements accepts any string as binding.source (it just
+    defaults a missing one to 'config'), so a 'live' or 'sensor' source --
+    anything a client sends that isn't literally 'config' -- hit the same
+    empty-if-block bug as an undrawable element type, just one level deeper:
+    the branch is taken, but its own inner `if` produced nothing.
+    """
+    element = {"type": "dynamic_text", "x": 0, "y": 0, "color": "#ffffff",
+               "binding": {"source": "live", "key": "temperature"}, **wrapper}
+    files = generate(element)          # must not raise ComposerInputError
+    assert "manager.py" in files
+    assert "non-config dynamic_text binding draws nothing" in files["manager.py"]
+
+
+def test_binding_source_cannot_break_out_of_the_comment_it_lands_in():
+    """binding.source used to be interpolated straight into a Python comment
+    (`pass  # dynamic_text binding_source "{{ el.binding_source }}" draws
+    nothing`) with no escaping. A source string carrying a newline closed the
+    comment, and text on the following line(s), indented to match, became a
+    real statement in the generated plugin -- CWE-94, and live once
+    composer_bp was registered (confirmed: this exact payload produces a
+    manager.py containing a clean, ast.parse-valid `import os` against the
+    pre-fix template). The comment is now a fixed literal that never
+    interpolates the value at all.
+    """
+    payload = "foo\n            import os\n            os.system('id')  # "
+    element = {"type": "dynamic_text", "x": 0, "y": 0, "color": "#ffffff",
+               "binding": {"source": payload, "key": "temperature"}}
+    files = generate(element)
+    src = files["manager.py"]
+    assert "import os" not in src
+    assert "os.system" not in src
+    assert "non-config dynamic_text binding draws nothing" in src
+    ast.parse(src)          # belt and braces: generate() already enforces this

test/test_composer_js_contracts.py

@@ -0,0 +1,263 @@
+"""Structural checks on the composer's JavaScript.
+
+There is no JS test runner in this repo, but three defects here are structural
+enough to assert on the parse tree, and each was a real bug:
+
+- Stroke widths inside _drawElement were left in canvas pixels while the
+  geometry around them scaled by SCALE, so at SCALE>1 every outline rendered
+  thinner than one LED pixel and the preview stopped matching the panel.
+- The `line` branch drew raw el.x0/el.y0/el.x1/el.y1, ignoring the anchor that
+  every other element type honours, so setting xAnchor moved everything except
+  lines -- and getBoundingBox had the same omission, leaving the hit box behind.
+- Four methods mutated exactly what _snapshot() serialises (metadata,
+  currentPreset) without calling it. _snapshot is the only caller of
+  _debouncedAutosave, so those changes were lost on reload and could not be
+  undone.
+"""
+import re
+from pathlib import Path
+
+import pytest
+
+tree_sitter = pytest.importorskip("tree_sitter")
+tree_sitter_javascript = pytest.importorskip("tree_sitter_javascript")
+
+JS_DIR = Path(__file__).resolve().parent.parent / "web_interface/static/v3/js/composer"
+CANVAS = JS_DIR / "composer-canvas.js"
+APP = JS_DIR / "composer-app.js"
+
+
+def _function_source(path: Path, name: str) -> str:
+    """Return the source of a top-level function declaration by name."""
+    src = path.read_bytes()
+    lang = tree_sitter.Language(tree_sitter_javascript.language())
+    tree = tree_sitter.Parser(lang).parse(src)
+    found = []
+
+    def walk(node):
+        if node.type == "function_declaration":
+            ident = node.child_by_field_name("name")
+            if ident is not None and src[ident.start_byte:ident.end_byte].decode() == name:
+                found.append(src[node.start_byte:node.end_byte].decode())
+        for c in node.children:
+            walk(c)
+
+    walk(tree.root_node)
+    assert found, f"{name} not found in {path.name}"
+    return found[0]
+
+
+def _method_source(path: Path, name: str) -> str:
+    """Return the source of a top-level object method by name."""
+    src = path.read_bytes()
+    lang = tree_sitter.Language(tree_sitter_javascript.language())
+    tree = tree_sitter.Parser(lang).parse(src)
+    found = []
+
+    def walk(node):
+        if node.type == "method_definition":
+            ident = node.child_by_field_name("name")
+            if ident is not None and src[ident.start_byte:ident.end_byte].decode() == name:
+                found.append(src[node.start_byte:node.end_byte].decode())
+        for c in node.children:
+            walk(c)
+
+    walk(tree.root_node)
+    assert found, f"{name} not found in {path.name}"
+    return found[0]
+
+
+def test_both_files_parse():
+    lang = tree_sitter.Language(tree_sitter_javascript.language())
+    parser = tree_sitter.Parser(lang)
+    for path in (CANVAS, APP):
+        tree = parser.parse(path.read_bytes())
+        errors = []
+
+        def walk(node):
+            if node.type == "ERROR" or node.is_missing:
+                errors.append(node.start_point[0] + 1)
+            for c in node.children:
+                walk(c)
+
+        walk(tree.root_node)
+        assert not errors, f"{path.name} has parse errors at lines {errors}"
+
+
+def test_element_strokes_scale_with_scale():
+    """No bare `ctx.lineWidth = 1` inside _drawElement.
+
+    Selection handles and the grid are drawn in canvas pixels deliberately and
+    live in other functions, so this is scoped to the element drawing routine.
+    """
+    body = _function_source(CANVAS, "_drawElement")
+    offenders = re.findall(r"ctx\.lineWidth\s*=\s*1\s*;", body)
+    assert not offenders, f"{len(offenders)} unscaled stroke width(s) in _drawElement"
+
+
+def test_line_branch_applies_the_anchor_offset():
+    """Scoped to _drawElement.
+
+    getBoundingBox has its own `case 'line': {` and appears first in the file,
+    so searching the whole text found *that* branch -- this assertion passed
+    with the draw branch's anchor offset removed. Verified: stripping it and
+    re-running gave 11/11 green.
+    """
+    body = _function_source(CANVAS, "_drawElement")
+    line_branch = body[body.index("case 'line': {"):]
+    line_branch = line_branch[:line_branch.index("case 'divider'")]
+    assert "ax - el.x0" in line_branch and "ay - el.y0" in line_branch, \
+        "line drawing ignores xAnchor/yAnchor"
+    assert "moveTo(el.x0 * s" not in line_branch, \
+        "line still drawn from unanchored endpoints"
+
+
+def test_line_bounding_box_applies_the_anchor_offset():
+    """The companion to the above: scoped to getBoundingBox specifically, so
+    the two tests cannot both be satisfied by the same branch."""
+    body = _function_source(CANVAS, "getBoundingBox")
+    box = body[body.index("case 'line'"):]
+    box = box[:box.index("case 'divider'")]
+    assert "ax - el.x0" in box, "line bounding box ignores the anchor"
+
+
+@pytest.mark.parametrize("method", [
+    "onBgColorChange",     # mutates metadata.bgColor
+    "setCustomSize",       # mutates currentPreset / MATRIX_W / MATRIX_H
+    "changePreset",        # mutates currentPreset / MATRIX_W / MATRIX_H
+    "applyPresetLabel",    # same, for sizes not in DISPLAY_PRESETS
+    "onColorChange",       # the one that was already fixed — keeps it fixed
+])
+def test_state_mutations_take_a_snapshot(method):
+    body = _method_source(APP, method)
+    assert "_snapshot()" in body, \
+        f"{method} changes snapshotted state without calling _snapshot()"
+    assert "isDirty = true" in body, f"{method} does not mark the design dirty"
+
+
+@pytest.mark.parametrize("method", ["changePreset", "applyPresetLabel"])
+def test_restore_path_stays_snapshot_free(method):
+    """_applyState and loadTemplate call these with {silent: true} while
+    restoring; snapshotting there would push restore steps onto the undo stack
+    and re-autosave the state just loaded."""
+    body = _method_source(APP, method)
+    assert "opts.silent" in body, f"{method} lost its silent guard"
+    snap = body.index("_snapshot()")
+    guard = body.index("!opts.silent")
+    assert guard < snap, f"{method} snapshots outside the !opts.silent guard"
+
+
+TEMPLATE_HTML = (Path(__file__).resolve().parent.parent
+                 / "web_interface/templates/v3/composer.html")
+
+#: The six toolbar buttons and the wrapper each must call.
+ALIGN_BUTTONS = ["alignLeft", "alignHCenter", "alignRight",
+                 "alignTop", "alignVCenter", "alignBottom"]
+
+
+def test_alignment_buttons_use_the_anchor_clearing_path():
+    """Two alignment implementations existed and the toolbar used the wrong one.
+
+    The legacy alignElement(dir) set el.x/el.y but left xAnchor/yAnchor in
+    place. resolveAnchor turns anchor='right' into `dim - val`, so "align left"
+    (el.x = 0) resolved to x = MATRIX_W -- the element jumped to the far right
+    edge instead. _alignElement clears the anchor first, so the stored value is
+    absolute, and it also updates el.x0/el.y0 so lines actually move.
+    """
+    html = TEMPLATE_HTML.read_text()
+    for wrapper in ALIGN_BUTTONS:
+        assert f"{wrapper}()" in html, f"toolbar does not call {wrapper}()"
+    assert not re.search(r"[^_]alignElement\(", html), \
+        "toolbar still calls the legacy alignElement()"
+
+
+def test_the_legacy_alignelement_is_gone():
+    """Leaving it in place invites the toolbar drifting back to it."""
+    src = APP.read_text()
+    assert not re.search(r"^\s{4}alignElement\(dir\)", src, re.M), \
+        "legacy alignElement(dir) still defined"
+
+
+def test_align_clears_the_anchor_and_moves_line_endpoints():
+    body = _method_source(APP, "_alignElement")
+    assert "xAnchor = null" in body and "yAnchor = null" in body, \
+        "_alignElement no longer clears the anchor, so aligning an anchored " \
+        "element resolves to the wrong edge"
+    assert "el.x0" in body and "el.y0" in body, \
+        "_alignElement no longer moves line endpoints"
+
+
+def test_align_translates_both_line_endpoints_not_just_the_start():
+    """Setting only x0 (or y0) left x1/y1 behind, so aligning a line changed
+    its shape instead of moving it -- e.g. a line from x0=20 to x1=50 aligned
+    right became x0=98, x1=50, stretching rather than translating it. Both
+    endpoints must move by the same delta."""
+    body = _method_source(APP, "_alignElement")
+    x_branch = body[body.index("if (axis === 'x')"):body.index("} else {")]
+    y_branch = body[body.index("} else {"):]
+    assert "el.x1" in x_branch, \
+        "_alignElement moves x0 but not x1 -- a line's shape changes, not its position"
+    assert "el.y1" in y_branch, \
+        "_alignElement moves y0 but not y1 -- a line's shape changes, not its position"
+
+
+def test_divider_stroke_is_centered_in_led_pixels():
+    """A 0.5 canvas-pixel offset (correct only at SCALE=1) was applied after
+    scaling instead of before it, so at SCALE>1 the stroke bled into the
+    preceding LED row/column instead of straddling its own."""
+    body = _function_source(CANVAS, "_drawElement")
+    divider_branch = body[body.index("case 'divider'"):]
+    divider_branch = divider_branch[:divider_branch.index("case 'pips'")]
+    assert "ay * s + 0.5" not in divider_branch, \
+        "divider still offsets by 0.5 canvas pixels after scaling"
+    assert "ax * s + 0.5" not in divider_branch, \
+        "divider still offsets by 0.5 canvas pixels after scaling"
+    assert "(ay + 0.5) * s" in divider_branch and "(ax + 0.5) * s" in divider_branch, \
+        "divider stroke is not centered within its LED pixel"
+
+
+def test_gauge_radii_are_clamped_to_zero():
+    """An imported design can carry a small gauge with a wide lineWidth --
+    width=1, height=1, lineWidth=3 sends a negative radius into
+    ctx.ellipse(), which throws IndexSizeError and aborts render() for every
+    element still to be drawn, not just the gauge."""
+    body = _function_source(CANVAS, "_drawElement")
+    gauge_branch = body[body.index("case 'gauge'"):]
+    assert re.search(r"Math\.max\(0,\s*rx\s*-\s*lwPx\s*/\s*2\)", gauge_branch), \
+        "gauge x-radius is not clamped to zero"
+    assert re.search(r"Math\.max\(0,\s*ry\s*-\s*lwPx\s*/\s*2\)", gauge_branch), \
+        "gauge y-radius is not clamped to zero"
+    assert "rx - lwPx / 2" not in re.sub(r"Math\.max\(0,\s*rx\s*-\s*lwPx\s*/\s*2\)", "", gauge_branch), \
+        "an unclamped gauge radius is still passed to ctx.ellipse()"
+
+
+#: ELEMENT_DEFAULTS in composer-canvas.js gives exactly these types a
+#: `binding` object; verified against that file rather than assumed.
+BOUND_TYPES = ["dynamic_text", "progress_bar", "countdown", "pips", "sparkline", "gauge"]
+
+
+def test_bound_types_constant_lists_every_bound_element_type():
+    """ELEMENT_DEFAULTS in composer-canvas.js gives dynamic_text, progress_bar,
+    countdown, pips, sparkline and gauge a `binding` object. If a type is added
+    to (or removed from) that list without updating APP's BOUND_TYPES, the
+    checks below drift out of sync silently -- this pins the two together."""
+    src = APP.read_text()
+    match = re.search(r"const BOUND_TYPES = \[([^\]]*)\];", src)
+    assert match, "BOUND_TYPES constant not found in composer-app.js"
+    declared = {t.strip().strip("'\"") for t in match.group(1).split(",") if t.strip()}
+    assert declared == set(BOUND_TYPES), \
+        f"BOUND_TYPES {declared} does not match the bound element types {set(BOUND_TYPES)}"
+
+
+@pytest.mark.parametrize("method", ["_isBound", "removeConfigVar", "_validateBeforeExport"])
+def test_binding_checks_cover_every_bound_element_type(method):
+    """Only dynamic_text and progress_bar were checked, so a countdown, pips,
+    sparkline or gauge element with an empty binding passed export validation
+    silently, and deleting a config var still used by one of them gave no
+    warning -- the plugin would read a now-missing key at runtime with no
+    indication why."""
+    body = _method_source(APP, method)
+    assert "BOUND_TYPES.includes(e.type)" in body, \
+        f"{method} does not check every bound element type via BOUND_TYPES"
+    assert "e.type === 'dynamic_text'" not in body, \
+        f"{method} still hardcodes only dynamic_text instead of BOUND_TYPES"

test/test_composer_path_containment.py

@@ -0,0 +1,225 @@
+"""A composer plugin id must never resolve outside the plugins directory.
+
+CodeQL reported sixteen high-severity py/path-injection alerts against
+web_interface/blueprints/composer.py: a request-supplied plugin_id reaching
+Path(plugins_dir) / plugin_id, which is then created, written to, deleted
+(shutil.rmtree) and read back.
+
+The id was already validated by an anchored regex, so every traversal payload
+was in fact rejected. What was missing was the guarantee living *with* the path
+building rather than in a regex several hundred lines away -- loosen that regex
+later and the traversal opens silently, with nothing at the filesystem boundary
+to catch it. _plugin_dir() closes that, and is the form static analysis can see.
+"""
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from web_interface.blueprints import composer as C  # noqa: E402
+
+#: Anything that has ever been used to climb out of a directory.
+TRAVERSAL = [
+    "../../etc/passwd", "..", ".", "a/../../etc", "good/../../..",
+    "/etc/passwd", "//etc/passwd", "a\\..\\..", "a%2f..%2f..",
+    "....//....//etc", "a/./../../etc", "~", "~root",
+    "plugin/../../../../../../etc/shadow",
+]
+
+#: Rejected for shape, not traversal -- but rejected all the same.
+MALFORMED = ["", "A-upper", "1-leading-digit", "-leading-dash", "has_underscore",
+             "has space", "has.dot", "a" * 64, "plugin\n", "plugin\n../../etc",
+             "\n", "plug\x00in"]
+
+
+@pytest.fixture
+def plugins_dir(tmp_path, monkeypatch):
+    base = tmp_path / "plugin-repos"
+    base.mkdir()
+    monkeypatch.setattr(C.composer_bp, "plugins_dir", str(base), raising=False)
+    return base
+
+
+@pytest.mark.parametrize("payload", TRAVERSAL)
+def test_traversal_payloads_are_refused(plugins_dir, payload):
+    assert C._plugin_dir(payload) is None
+
+
+@pytest.mark.parametrize("payload", MALFORMED)
+def test_malformed_ids_are_refused(plugins_dir, payload):
+    assert C._plugin_dir(payload) is None
+
+
+@pytest.mark.parametrize("payload", ["a", "my-plugin", "x9", "a" * 63])
+def test_valid_ids_resolve_inside_the_base(plugins_dir, payload):
+    resolved = C._plugin_dir(payload)
+    assert resolved is not None, f"{payload!r} was rejected but is valid"
+    assert resolved.parent == plugins_dir.resolve(), (
+        f"{payload!r} resolved to {resolved}, outside {plugins_dir}")
+
+
+def test_no_payload_can_escape_even_if_the_regex_is_loosened(plugins_dir, monkeypatch):
+    """The containment check must stand on its own.
+
+    This is the whole point of resolving at the filesystem boundary: if the id
+    pattern is ever relaxed, traversal must still be impossible. Replace the
+    regex with one that permits slashes and dots, then re-run the payloads.
+    """
+    import re
+    monkeypatch.setattr(C, "_PLUGIN_ID_RE", re.compile(r"\A[\w./\\~-]+\Z"))
+    import os
+    escaped = []
+    base = os.path.realpath(str(plugins_dir))
+    for payload in TRAVERSAL:
+        resolved = C._plugin_dir(payload)
+        if resolved is None:
+            continue
+        real = os.path.realpath(str(resolved))
+        if real != base and os.path.commonpath([base, real]) != base:
+            escaped.append((payload, real))
+    assert not escaped, f"these escaped the base with a loosened regex: {escaped}"
+
+
+def test_a_sibling_directory_with_a_shared_prefix_is_not_inside(tmp_path, monkeypatch):
+    """commonpath, not startswith.
+
+    "/x/plugins-evil" starts with "/x/plugins" but is a different directory, so
+    a prefix test would accept it.
+    """
+    base = tmp_path / "plugins"
+    base.mkdir()
+    (tmp_path / "plugins-evil").mkdir()
+    monkeypatch.setattr(C.composer_bp, "plugins_dir", str(base), raising=False)
+    import re
+    # Neutralise the two layers in front so this exercises the containment
+    # check itself; otherwise secure_filename rejects the payload first and a
+    # startswith regression would go unnoticed here.
+    monkeypatch.setattr(C, "_PLUGIN_ID_RE", re.compile(r"\A[\w./\\~-]+\Z"))
+    monkeypatch.setattr(C, "secure_filename", lambda v: v)
+    assert C._plugin_dir("../plugins-evil") is None
+
+
+def test_containment_still_holds_if_the_sanitiser_is_defeated(plugins_dir, monkeypatch):
+    """Each layer is tested on its own, not just the stack.
+
+    secure_filename's equality guard rejects every traversal payload before the
+    containment check sees it, so removing containment does not fail the other
+    tests -- which would make it look load-bearing when it is not. Neutralise
+    the regex *and* the sanitiser, and the realpath/commonpath check must still
+    refuse everything on its own.
+    """
+    import re
+    monkeypatch.setattr(C, "_PLUGIN_ID_RE", re.compile(r"\A[\w./\\~-]+\Z"))
+    monkeypatch.setattr(C, "secure_filename", lambda v: v)
+    import os
+    base = os.path.realpath(str(plugins_dir))
+    escaped = []
+    for payload in TRAVERSAL:
+        resolved = C._plugin_dir(payload)
+        if resolved is None:
+            continue
+        real = os.path.realpath(str(resolved))
+        # Inside the base is fine -- "...." and "~" are ordinary directory
+        # names on Linux, so they are not escapes. What must never happen is
+        # landing outside the base, or on the base itself: install() rmtrees
+        # its target, so the plugins root resolving to a "plugin" would wipe
+        # every installed plugin.
+        if real == base or os.path.commonpath([base, real]) != base:
+            escaped.append((payload, real))
+    assert not escaped, f"containment alone let these through: {escaped}"
+
+
+def test_secure_filename_never_rewrites_an_accepted_id(plugins_dir):
+    """The sanitiser must be a no-op on everything the regex accepts.
+
+    If secure_filename ever altered an accepted id, _plugin_dir would resolve
+    to a *different* plugin's directory than the caller asked for -- a silent
+    redirect, which is worse than a refusal. The guard turns that into a
+    refusal; this proves the guard never has to fire in practice.
+    """
+    import random
+    from werkzeug.utils import secure_filename
+    random.seed(1)
+    alphabet = "abcdefghijklmnopqrstuvwxyz0123456789-"
+    altered = []
+    for _ in range(2000):
+        n = random.randint(1, 63)
+        cand = random.choice("abcdefghijklmnopqrstuvwxyz") + "".join(
+            random.choice(alphabet) for _ in range(n - 1))
+        if C._PLUGIN_ID_RE.match(cand) and secure_filename(cand) != cand:
+            altered.append((cand, secure_filename(cand)))
+    assert not altered, f"secure_filename rewrote accepted ids: {altered[:5]}"
+
+
+def test_a_trailing_newline_is_not_a_valid_id():
+    r"""Python's `$` also matches before a trailing newline, so the original
+    `^...$` accepted "myplugin\n" and would have created a directory whose
+    name ends in one. \Z does not."""
+    assert C._PLUGIN_ID_RE.match("myplugin") is not None
+    assert C._PLUGIN_ID_RE.match("myplugin\n") is None
+
+
+# --- font serving -----------------------------------------------------------
+
+FONT_TRAVERSAL = [
+    "../../../etc/passwd", "../config/config.json", "..%2f..%2fetc%2fpasswd",
+    "PressStart2P-Regular.ttf/../../../etc/passwd", "/etc/passwd", "",
+    "PressStart2P-Regular.TTF",          # case differs -> not the allowlisted name
+    "PressStart2P-Regular.ttf ",         # trailing space
+]
+
+
+def test_serve_font_refuses_a_file_that_exists_but_is_not_allowlisted(monkeypatch, tmp_path):
+    """The allowlist must be what refuses it, not a missing file.
+
+    Asserting 404 on traversal payloads proves nothing here: Flask's router
+    will not match a path segment containing '/', and everything else 404s
+    simply because no such file exists. Put a real, readable file next to the
+    fonts and confirm it is still refused -- that is the allowlist working.
+    """
+    fonts = tmp_path / "assets" / "fonts"
+    fonts.mkdir(parents=True)
+    (fonts / "id_rsa.ttf").write_bytes(b"PRIVATE KEY")
+    monkeypatch.setattr(C.composer_bp, "project_root", str(tmp_path), raising=False)
+    app = __import__("flask").Flask(__name__)
+    app.register_blueprint(C.composer_bp)
+    with app.test_client() as client:
+        resp = client.get("/api/fonts/id_rsa.ttf")
+    assert resp.status_code == 404, (
+        "a readable non-allowlisted file was served; the allowlist is not gating")
+    assert b"PRIVATE KEY" not in resp.data
+
+
+@pytest.mark.parametrize("payload", FONT_TRAVERSAL)
+def test_serve_font_refuses_anything_not_allowlisted(payload, monkeypatch, tmp_path):
+    """The name reaching the filesystem must come from the allowlist constant.
+
+    _ALLOWED_FONTS gates this endpoint, so nothing here was ever exploitable.
+    Building the path from the matched constant rather than the request value
+    is what makes that provable -- and it is why CodeQL reported two
+    high-severity py/path-injection alerts on an endpoint that was already
+    safe.
+    """
+    monkeypatch.setattr(C.composer_bp, "project_root", str(tmp_path), raising=False)
+    app = C.composer_bp.name and __import__("flask").Flask(__name__)
+    app.register_blueprint(C.composer_bp)
+    with app.test_client() as client:
+        resp = client.get(f"/api/fonts/{payload}")
+    assert resp.status_code in (404, 405, 308), (
+        f"{payload!r} was not refused (status {resp.status_code})")
+
+
+def test_serve_font_still_serves_each_allowlisted_font(monkeypatch, tmp_path):
+    fonts = tmp_path / "assets" / "fonts"
+    fonts.mkdir(parents=True)
+    monkeypatch.setattr(C.composer_bp, "project_root", str(tmp_path), raising=False)
+    app = __import__("flask").Flask(__name__)
+    app.register_blueprint(C.composer_bp)
+    for name in C._ALLOWED_FONTS:
+        (fonts / name).write_bytes(b"\x00\x01ttf")
+        with app.test_client() as client:
+            resp = client.get(f"/api/fonts/{name}")
+        assert resp.status_code == 200, f"{name} should be served, got {resp.status_code}"

test/test_config_helper.py

@@ -0,0 +1,253 @@
+"""
+Tests for src/common/config_helper.py — pins the ConfigHelper contract.
+
+Covers: load/save round trips (missing/malformed files return {} rather
+than raising, non-ASCII preserved via ensure_ascii=False, top-level JSON
+lists returned as-is), dot-notation get/set including the silent-failure
+contract when an intermediate key holds a non-dict, merge_configs deep
+semantics with NO aliasing of the base config (the fixed bug — the old
+shallow copy let mutations of the merged result leak into base's nested
+dicts), simplified schema validation including the caught-TypeError path
+when a schema 'type' is given as a string, plugin config key conventions
+('{plugin_id}_config', enabled defaults True), and required-key checks
+where a key present with value None counts as present.
+"""
+
+import json
+
+import pytest
+
+from src.common.config_helper import ConfigHelper
+
+
+@pytest.fixture
+def helper():
+    return ConfigHelper()
+
+
+class TestLoadConfig:
+    def test_missing_file_returns_empty_dict(self, helper, tmp_path):
+        assert helper.load_config(tmp_path / "nope.json") == {}
+
+    def test_malformed_json_returns_empty_dict(self, helper, tmp_path):
+        path = tmp_path / "bad.json"
+        path.write_text("{ this is not json", encoding="utf-8")
+        assert helper.load_config(path) == {}
+
+    def test_top_level_list_returned_as_is(self, helper, tmp_path):
+        # load_config does not enforce a dict shape: a JSON list comes
+        # straight back. Pinned as a characterization of current behavior.
+        path = tmp_path / "list.json"
+        path.write_text("[1, 2, 3]", encoding="utf-8")
+        assert helper.load_config(path) == [1, 2, 3]
+
+
+class TestSaveConfig:
+    def test_round_trip(self, helper, tmp_path):
+        path = tmp_path / "config.json"
+        config = {'display': {'hardware': {'rows': 32}}, 'timezone': 'UTC'}
+        assert helper.save_config(config, path) is True
+        assert helper.load_config(path) == config
+
+    def test_creates_parent_directories(self, helper, tmp_path):
+        path = tmp_path / "deep" / "nested" / "config.json"
+        assert helper.save_config({'a': 1}, path) is True
+        assert path.exists()
+        assert helper.load_config(path) == {'a': 1}
+
+    def test_non_ascii_survives_round_trip(self, helper, tmp_path):
+        path = tmp_path / "config.json"
+        config = {'city': 'Zürich', 'note': 'météo ☀'}
+        assert helper.save_config(config, path) is True
+        assert helper.load_config(path) == config
+        # ensure_ascii=False: characters are written raw, not \u-escaped
+        assert 'Zürich' in path.read_text(encoding='utf-8')
+
+    def test_directory_path_returns_false_not_raise(self, helper, tmp_path):
+        assert helper.save_config({'a': 1}, tmp_path) is False
+
+
+class TestGetConfigValue:
+    def test_dot_notation_hit(self, helper):
+        config = {'display': {'hardware': {'rows': 32}}}
+        assert helper.get_config_value(config, 'display.hardware.rows') == 32
+
+    def test_missing_returns_default(self, helper):
+        sentinel = object()
+        assert helper.get_config_value({}, 'display.rows', default=sentinel) is sentinel
+
+    def test_intermediate_non_dict_returns_default(self, helper):
+        config = {'display': 'not-a-dict'}
+        assert helper.get_config_value(config, 'display.hardware.rows', default=64) == 64
+
+    def test_required_missing_raises_keyerror(self, helper):
+        with pytest.raises(KeyError):
+            helper.get_config_value({}, 'display.rows', required=True)
+
+
+class TestSetConfigValue:
+    def test_sets_top_level(self, helper):
+        config = {}
+        helper.set_config_value(config, 'timezone', 'UTC')
+        assert config == {'timezone': 'UTC'}
+
+    def test_auto_creates_intermediates(self, helper):
+        config = {}
+        helper.set_config_value(config, 'display.hardware.rows', 32)
+        assert config == {'display': {'hardware': {'rows': 32}}}
+
+    def test_silent_failure_on_non_dict_intermediate(self, helper):
+        # 'a' exists but holds an int; the assignment attempt raises
+        # TypeError internally, which set_config_value swallows and logs.
+        # The config is left unchanged — pinned silent-failure contract.
+        config = {'a': 5}
+        helper.set_config_value(config, 'a.b', 1)
+        assert config == {'a': 5}
+
+
+class TestMergeConfigs:
+    def test_nested_dicts_merge_recursively(self, helper):
+        base = {'display': {'rows': 32, 'cols': 64}, 'timezone': 'UTC'}
+        override = {'display': {'cols': 128, 'brightness': 90}}
+        merged = helper.merge_configs(base, override)
+        assert merged == {
+            'display': {'rows': 32, 'cols': 128, 'brightness': 90},
+            'timezone': 'UTC',
+        }
+
+    def test_scalar_override_wins_over_dict(self, helper):
+        merged = helper.merge_configs({'display': {'rows': 32}}, {'display': 7})
+        assert merged['display'] == 7
+
+    def test_dict_override_wins_over_scalar(self, helper):
+        merged = helper.merge_configs({'display': 7}, {'display': {'rows': 32}})
+        assert merged['display'] == {'rows': 32}
+
+    def test_no_aliasing_of_base(self, helper):
+        # Post-fix: merge deep-copies base, so mutating the result never
+        # leaks back into the caller's base config.
+        base = {'display': {'x': 1}}
+        merged = helper.merge_configs(base, {})
+        assert merged['display'] is not base['display']
+        merged['display']['x'] = 99
+        assert base['display']['x'] == 1
+
+    def test_inputs_unchanged(self, helper):
+        base = {'a': {'b': 1}}
+        override = {'a': {'c': 2}}
+        helper.merge_configs(base, override)
+        assert base == {'a': {'b': 1}}
+        assert override == {'a': {'c': 2}}
+
+    def test_no_aliasing_of_override_values(self, helper):
+        # The non-recursive branch must deep-copy the override value too:
+        # mutating a merged-in list or dict must not reach back into
+        # override_config.
+        override = {'teams': ['A', 'B'], 'nested': {'x': [1]}}
+        merged = helper.merge_configs({}, override)
+        merged['teams'].append('C')
+        merged['nested']['x'].append(2)
+        assert override == {'teams': ['A', 'B'], 'nested': {'x': [1]}}
+
+
+class TestValidateConfig:
+    def test_no_schema_dict_is_valid(self, helper):
+        assert helper.validate_config({'a': 1}) is True
+
+    def test_no_schema_list_is_invalid(self, helper):
+        assert helper.validate_config([1, 2]) is False
+
+    def test_required_key_missing_is_invalid(self, helper):
+        schema = {'rows': {'required': True, 'type': int}}
+        assert helper.validate_config({}, schema) is False
+
+    def test_optional_key_missing_is_valid(self, helper):
+        schema = {'rows': {'required': False, 'type': int}}
+        assert helper.validate_config({}, schema) is True
+
+    def test_wrong_type_is_invalid(self, helper):
+        schema = {'rows': {'type': int}}
+        assert helper.validate_config({'rows': 'thirty-two'}, schema) is False
+        assert helper.validate_config({'rows': 32}, schema) is True
+
+    def test_allowed_values_violation_is_invalid(self, helper):
+        schema = {'mode': {'allowed_values': ['clock', 'weather']}}
+        assert helper.validate_config({'mode': 'stocks'}, schema) is False
+        assert helper.validate_config({'mode': 'clock'}, schema) is True
+
+    def test_string_type_in_schema_is_invalid_via_typeerror(self, helper):
+        # 'type' given as the STRING "int" makes isinstance() raise
+        # TypeError; validate_config catches it and returns False rather
+        # than raising. Pinned characterization.
+        schema = {'rows': {'type': 'int'}}
+        assert helper.validate_config({'rows': 32}, schema) is False
+
+
+class TestPluginConfigHelpers:
+    def test_get_plugin_config_uses_suffixed_key(self, helper):
+        plugin_cfg = {'enabled': True, 'display_duration': 30}
+        assert helper.get_plugin_config({'clock_config': plugin_cfg}, 'clock') == plugin_cfg
+
+    def test_get_plugin_config_bare_id_key_not_found(self, helper):
+        # Only '{plugin_id}_config' is consulted — a bare 'clock' section
+        # is invisible to this helper. Pinned key contract.
+        assert helper.get_plugin_config({'clock': {'enabled': True}}, 'clock') == {}
+
+    def test_create_default_config_wraps_in_suffixed_key(self, helper):
+        defaults = {'enabled': True}
+        assert helper.create_default_config('clock', defaults) == {'clock_config': defaults}
+
+    def test_is_plugin_enabled_defaults_true_for_unknown(self, helper):
+        assert helper.is_plugin_enabled({}, 'clock') is True
+
+    def test_is_plugin_enabled_false_when_disabled(self, helper):
+        config = {'clock_config': {'enabled': False}}
+        assert helper.is_plugin_enabled(config, 'clock') is False
+
+    def test_is_plugin_enabled_ignores_bare_id_key(self, helper):
+        # Disabled under the wrong key -> still reported enabled (default).
+        config = {'clock': {'enabled': False}}
+        assert helper.is_plugin_enabled(config, 'clock') is True
+
+
+class TestSportsAndDisplayHelpers:
+    def test_get_display_config(self, helper):
+        display = {'hardware': {'rows': 32}}
+        assert helper.get_display_config({'display': display}) == display
+        assert helper.get_display_config({}) == {}
+
+    def test_get_sports_config_uses_scoreboard_suffix(self, helper):
+        sport_cfg = {'favorite_teams': ['TB']}
+        config = {'football_scoreboard': sport_cfg}
+        assert helper.get_sports_config(config, 'football') == sport_cfg
+        assert helper.get_sports_config(config, 'hockey') == {}
+
+    def test_get_favorite_teams(self, helper):
+        config = {'football_scoreboard': {'favorite_teams': ['TB', 'DAL']}}
+        assert helper.get_favorite_teams(config, 'football') == ['TB', 'DAL']
+        assert helper.get_favorite_teams({}, 'football') == []
+
+    def test_get_display_modes(self, helper):
+        modes = {'live': True, 'recent': False}
+        config = {'football_scoreboard': {'display_modes': modes}}
+        assert helper.get_display_modes(config, 'football') == modes
+        assert helper.get_display_modes({}, 'football') == {}
+
+
+class TestValidateRequiredKeys:
+    def test_returns_missing_subset(self, helper):
+        config = {'a': 1, 'c': {'d': 2}}
+        missing = helper.validate_required_keys(config, ['a', 'b', 'c.d', 'c.e'])
+        assert missing == ['b', 'c.e']
+
+    def test_dot_notation_present(self, helper):
+        config = {'display': {'hardware': {'rows': 32}}}
+        assert helper.validate_required_keys(config, ['display.hardware.rows']) == []
+
+    def test_empty_requirements(self, helper):
+        assert helper.validate_required_keys({'a': 1}, []) == []
+
+    def test_present_with_none_counts_as_present(self, helper):
+        # _has_key checks key membership, not truthiness — a key set to
+        # None is NOT reported missing. Pinned semantics.
+        assert helper.validate_required_keys({'a': None}, ['a']) == []

test/test_config_main_redacts_secrets.py

@@ -0,0 +1,143 @@
+"""GET /config/main must not hand out credentials.
+
+The endpoint returned the raw config to anyone who could reach the port, and
+this web interface has no authentication of any kind. Measured against a live
+rig, an unauthenticated request returned:
+
+    github.api_token                40 chars
+    incoming-packages.ha_token     183 chars
+    jellyfin-now-playing.api_key    32 chars
+    ledmatrix-weather.api_key       32 chars
+    on-air.mqtt_password             8 chars
+    youtube.api_key                 20 chars
+    youtube-stats.api_key           39 chars
+
+A GitHub token and a Home Assistant long-lived token among them.
+
+The x-secret masking the plugin config endpoints use does not apply here: this
+endpoint never consults a schema, and core keys such as github.api_token have
+no schema to carry the marker. Several of those fields *are* tagged x-secret in
+their plugin's schema and were still returned in full, which is what makes the
+schema route the wrong one to rely on for this endpoint.
+
+Matching on field name is blunt. For a whole-config dump it is the right
+default: anything named like a credential should not leave the process, and a
+new plugin that adds a differently-shaped secret is covered without anyone
+remembering to tag it.
+"""
+import pytest
+
+from web_interface.blueprints.api_v3 import (
+    _looks_like_a_credential,
+    _redact_credentials,
+)
+
+
+@pytest.mark.parametrize("name", [
+    "password", "mqtt_password", "opensky_password", "passwd",
+    "api_key", "apikey", "API_KEY", "flightaware_api_key",
+    "token", "ha_token", "api_token", "access_token",
+    "secret", "client_secret", "spotify_client_secret",
+    "access_key", "private_key",
+])
+def test_credential_names_are_recognised(name):
+    assert _looks_like_a_credential(name)
+
+
+@pytest.mark.parametrize("name", [
+    "timezone", "city", "brightness", "enabled", "update_interval",
+    "favorite_teams", "display_duration", "keyword",
+])
+def test_ordinary_names_are_left_alone(name):
+    assert not _looks_like_a_credential(name)
+
+
+def test_the_measured_leak_is_closed():
+    """The exact shape taken off the rig."""
+    config = {
+        "github": {"api_token": "ghp_" + "x" * 36},
+        "incoming-packages": {"ha_token": "y" * 183, "enabled": True},
+        "jellyfin-now-playing": {"api_key": "z" * 32},
+        "on-air": {"mqtt_password": "hunter22"},
+        "youtube": {"api_key": "k" * 20},
+        "timezone": "America/New_York",
+    }
+    out = _redact_credentials(config)
+    assert out["github"]["api_token"] == ""
+    assert out["incoming-packages"]["ha_token"] == ""
+    assert out["jellyfin-now-playing"]["api_key"] == ""
+    assert out["on-air"]["mqtt_password"] == ""
+    assert out["youtube"]["api_key"] == ""
+    # Everything else survives, or the config editor breaks.
+    assert out["timezone"] == "America/New_York"
+    assert out["incoming-packages"]["enabled"] is True
+
+
+def test_nested_and_listed_credentials_are_reached():
+    config = {"a": {"b": {"c": {"password": "p"}}},
+              "feeds": [{"name": "x", "api_key": "k"}, {"name": "y"}]}
+    out = _redact_credentials(config)
+    assert out["a"]["b"]["c"]["password"] == ""
+    assert out["feeds"][0]["api_key"] == ""
+    assert out["feeds"][0]["name"] == "x"
+
+
+def test_the_original_is_not_mutated():
+    """The caller holds the live config; redaction must not edit it in place."""
+    config = {"github": {"api_token": "keepme"}}
+    _redact_credentials(config)
+    assert config["github"]["api_token"] == "keepme"
+
+
+def test_a_credential_shaped_container_is_still_walked():
+    """`secrets: {...}` is a section name, not a value to blank."""
+    config = {"secrets": {"api_key": "k", "note": "keep"}}
+    out = _redact_credentials(config)
+    assert out["secrets"]["api_key"] == ""
+    assert out["secrets"]["note"] == "keep"
+
+
+def test_non_dict_input_passes_through():
+    assert _redact_credentials("plain") == "plain"
+    assert _redact_credentials(7) == 7
+    assert _redact_credentials(None) is None
+
+
+def test_the_endpoint_itself_redacts():
+    """Through the view function, not the helper.
+
+    The helper tests above all passed with the route still returning
+    `config` -- reverting the one line that calls the redactor changed
+    nothing, because nothing exercised the route. A property asserted on a
+    helper is not a property asserted on the endpoint, and it is the endpoint
+    that is exposed to the network.
+    """
+    import json as _json
+    from unittest.mock import MagicMock
+
+    import flask
+
+    from web_interface.blueprints import api_v3 as mod
+
+    raw = {"github": {"api_token": "ghp_secret_value"},
+           "timezone": "America/New_York"}
+
+    manager = MagicMock()
+    manager.load_config.return_value = raw
+    previous = getattr(mod.api_v3, "config_manager", None)
+    mod.api_v3.config_manager = manager
+
+    app = flask.Flask(__name__)
+    try:
+        with app.test_request_context("/config/main"):
+            response = mod.get_main_config()
+        payload = response.get_json() if hasattr(response, "get_json") else _json.loads(response[0].data)
+    finally:
+        mod.api_v3.config_manager = previous
+
+    data = payload["data"]
+    assert data["github"]["api_token"] == "", (
+        "the endpoint returned the token; the redactor is not wired in")
+    assert data["timezone"] == "America/New_York"
+    # And the config the manager handed over is untouched.
+    assert raw["github"]["api_token"] == "ghp_secret_value"

test/test_config_manager_secrets.py

@@ -177,3 +177,159 @@ class TestLoadFastPath:
         os.utime(config_file, ns=(1_000_000_000, 1_000_000_000))
 
         assert manager.load_config()["timezone"] == "AAA"  # stale, by design
+
+
+class TestArraySecretStripAndMerge:
+    """Array-item secrets round-trip (parallel-placeholder lists).
+
+    secret_helpers.separate_secrets emits array secrets as a list parallel
+    to the regular list, with {} for items that carry no secrets. Strip
+    must remove the secret fields from config.json while preserving item
+    indices; load must merge them back into the right items. The regular
+    list's length is authoritative in both directions.
+    """
+
+    def test_strip_removes_array_item_secrets_keeps_indices(self, tmp_path):
+        manager = make_manager(tmp_path)
+        data = {"plugin": {"accounts": [
+            {"name": "a", "token": "ta"},
+            {"name": "b"},
+        ]}}
+        secrets = {"plugin": {"accounts": [{"token": "ta"}, {}]}}
+        stripped = manager._strip_secrets_recursive(data, secrets)
+        assert stripped == {"plugin": {"accounts": [{"name": "a"}, {"name": "b"}]}}
+
+    def test_strip_keeps_all_placeholder_items(self, tmp_path):
+        # Even when every item strips to nothing extra, the list survives
+        # with its indices — required for merge-on-load alignment.
+        manager = make_manager(tmp_path)
+        data = {"accounts": [{"token": "t1"}, {"token": "t2"}]}
+        secrets = {"accounts": [{"token": "t1"}, {"token": "t2"}]}
+        stripped = manager._strip_secrets_recursive(data, secrets)
+        assert stripped == {"accounts": [{}, {}]}
+
+    def test_strip_whole_scalar_array_secret_drops_key(self, tmp_path):
+        # A list of secret scalars is a whole-key secret, not the parallel
+        # shape — the key must vanish from config.json entirely.
+        manager = make_manager(tmp_path)
+        data = {"recovery_codes": ["a", "b"], "city": "Austin"}
+        secrets = {"recovery_codes": ["a", "b"]}
+        stripped = manager._strip_secrets_recursive(data, secrets)
+        assert stripped == {"city": "Austin"}
+
+    def test_strip_shape_mismatch_drops_key(self, tmp_path):
+        # Conservative contract: if the shapes disagree, never leak.
+        manager = make_manager(tmp_path)
+        data = {"accounts": {"name": "not-a-list"}}
+        secrets = {"accounts": [{"token": "t"}]}
+        stripped = manager._strip_secrets_recursive(data, secrets)
+        assert stripped == {}
+
+    def test_strip_ignores_extra_secrets_entries(self, tmp_path):
+        # Regular list length is authoritative: a user deleted an item.
+        manager = make_manager(tmp_path)
+        data = {"accounts": [{"name": "a", "token": "ta"}]}
+        secrets = {"accounts": [{"token": "ta"}, {"token": "tb"}]}
+        stripped = manager._strip_secrets_recursive(data, secrets)
+        assert stripped == {"accounts": [{"name": "a"}]}
+
+    def test_merge_restores_array_item_secrets(self, tmp_path):
+        manager = make_manager(tmp_path)
+        target = {"accounts": [{"name": "a"}, {"name": "b"}]}
+        manager._deep_merge(target, {"accounts": [{"token": "ta"}, {}]})
+        assert target == {"accounts": [
+            {"name": "a", "token": "ta"},
+            {"name": "b"},
+        ]}
+
+    def test_merge_ignores_extra_secrets_entries_with_warning(self, tmp_path, caplog):
+        manager = make_manager(tmp_path)
+        target = {"accounts": [{"name": "a"}]}
+        with caplog.at_level("WARNING"):
+            manager._deep_merge(
+                target, {"accounts": [{"token": "ta"}, {"token": "ghost"}]})
+        assert target == {"accounts": [{"name": "a", "token": "ta"}]}
+        assert any("longer than the config list" in r.message for r in caplog.records)
+
+    def test_merge_non_dict_item_replaced_by_secret(self, tmp_path):
+        # Shape drift inside the list: the secret wins for that index.
+        manager = make_manager(tmp_path)
+        target = {"accounts": ["oddball", {"name": "b"}]}
+        manager._deep_merge(target, {"accounts": [{"token": "ta"}, {}]})
+        assert target == {"accounts": [{"token": "ta"}, {"name": "b"}]}
+
+    def test_merge_whole_scalar_array_still_replaces(self, tmp_path):
+        # Legacy behavior preserved: a non-parallel list replaces wholesale.
+        manager = make_manager(tmp_path)
+        target = {"recovery_codes": ["old"]}
+        manager._deep_merge(target, {"recovery_codes": ["new1", "new2"]})
+        assert target == {"recovery_codes": ["new1", "new2"]}
+
+    def test_full_save_load_round_trip(self, tmp_path):
+        # End to end on real files: save strips array secrets out of
+        # config.json; load merges them back into the right items.
+        manager = make_manager(
+            tmp_path,
+            config={"plugin": {"accounts": [
+                {"name": "a", "token": "s3cret-a"},
+                {"name": "b", "token": "s3cret-b"},
+            ]}},
+            secrets={"plugin": {"accounts": [
+                {"token": "s3cret-a"}, {"token": "s3cret-b"},
+            ]}},
+        )
+        loaded = manager.load_config()
+        assert loaded["plugin"]["accounts"][0]["token"] == "s3cret-a"
+
+        manager.save_config(loaded)
+
+        raw = (tmp_path / "config.json").read_text()
+        assert "s3cret" not in raw
+        on_disk = json.loads(raw)
+        assert on_disk["plugin"]["accounts"] == [{"name": "a"}, {"name": "b"}]
+
+        # A fresh manager (constructed directly — make_manager would
+        # overwrite the just-saved config.json) re-merges from the secrets
+        # file on load.
+        fresh = ConfigManager(config_path=str(tmp_path / "config.json"),
+                              secrets_path=str(tmp_path / "config_secrets.json"))
+        fresh.template_path = str(tmp_path / "no-template.json")
+        reloaded = fresh.load_config()
+        assert reloaded["plugin"]["accounts"] == [
+            {"name": "a", "token": "s3cret-a"},
+            {"name": "b", "token": "s3cret-b"},
+        ]
+
+    def test_whole_item_secret_list_never_leaks_values(self, tmp_path):
+        # When the ENTIRE array item is secret (schema marks both key[]
+        # and key[].field), separate_secrets stores the full item dicts in
+        # the secrets file. That shape also matches the parallel-list
+        # discriminator — which is safe: strip drops every leaf key that
+        # appears in the secret item, so only empty {} skeletons (item
+        # count, no values) can reach config.json, and merge-on-load
+        # restores the full items from those skeletons.
+        from src.web_interface.secret_helpers import (
+            find_secret_fields, separate_secrets)
+        schema_props = {"accounts": {
+            "type": "array",
+            "items": {"type": "object", "x-secret": True, "properties": {
+                "id": {"type": "string"},
+                "token": {"type": "string", "x-secret": True},
+            }},
+        }}
+        paths = find_secret_fields(schema_props)
+        assert paths == {"accounts[]", "accounts[].token"}
+        full = {"accounts": [{"id": "i1", "token": "s3cret-a"},
+                             {"id": "i2", "token": "s3cret-b"}]}
+        _, secrets = separate_secrets(full, paths)
+        assert secrets == full  # whole items are secret
+
+        manager = make_manager(tmp_path)
+        stripped = manager._strip_secrets_recursive(full, secrets)
+        assert stripped == {"accounts": [{}, {}]}
+
+        raw = json.dumps(stripped)
+        assert "s3cret" not in raw and "i1" not in raw
+
+        manager._deep_merge(stripped, secrets)
+        assert stripped == full  # round trip restores the items

test/test_core_version_freshness.py

@@ -0,0 +1,122 @@
+"""The gate must read the core version from disk, not from its own import.
+
+`from src import __version__` binds whatever the process loaded at start. The
+web UI is a long-lived service of its own (ledmatrix-web.service), and updating
+the core replaces files on disk without restarting it -- the update route says
+so and asks the user to restart, but its prompt named only the *display*
+service, so a user who followed it left the web process holding the old number.
+
+The plugin store's gate lives in that web process, so being stale by exactly
+one release is the case that bites: every plugin flooring on the release you
+just installed is refused, with a message blaming a core version that is
+already correct on disk.
+
+Observed on hardware: after updating a rig to 3.3.0 and restarting only the
+display service, all eight sports scoreboards were refused with
+"supports LEDMatrix >=3.3.0, but this system is running 3.2.0" while
+src/__init__.py on that machine read 3.3.0.
+"""
+
+import importlib
+import sys
+
+import pytest
+
+from src.plugin_system import compatibility
+
+
+class TestCurrentCoreVersion:
+    def test_it_reads_the_file_rather_than_the_imported_value(self, monkeypatch, tmp_path):
+        # Simulate a process whose import predates the update: the module
+        # object says 3.2.0 while the file on disk says 3.3.0.
+        import src
+        monkeypatch.setattr(src, "__version__", "3.2.0")
+        fake = tmp_path / "__init__.py"
+        fake.write_text('__version__ = "3.3.0"\n', encoding="utf-8")
+        monkeypatch.setattr(compatibility, "_VERSION_FILE", fake)
+        assert compatibility.current_core_version() == "3.3.0"
+
+    def test_it_matches_the_real_file_by_default(self):
+        import src
+        assert compatibility.current_core_version() == src.__version__
+
+    @pytest.mark.parametrize("body", [
+        "__version__ = '3.4.1'\n",
+        '__version__="3.4.1"\n',
+        '"""doc"""\n\n__version__ = "3.4.1"  # trailing comment\n',
+    ])
+    def test_it_tolerates_the_ways_that_line_gets_written(self, monkeypatch, tmp_path, body):
+        fake = tmp_path / "__init__.py"
+        fake.write_text(body, encoding="utf-8")
+        monkeypatch.setattr(compatibility, "_VERSION_FILE", fake)
+        assert compatibility.current_core_version() == "3.4.1"
+
+    def test_a_missing_file_falls_back_to_the_import(self, monkeypatch, tmp_path):
+        # Never worse than before: an unreadable file returns what the old
+        # code would have returned.
+        import src
+        monkeypatch.setattr(src, "__version__", "3.2.0")
+        monkeypatch.setattr(compatibility, "_VERSION_FILE", tmp_path / "gone.py")
+        assert compatibility.current_core_version() == "3.2.0"
+
+    def test_a_file_without_the_line_falls_back(self, monkeypatch, tmp_path):
+        import src
+        monkeypatch.setattr(src, "__version__", "3.2.0")
+        fake = tmp_path / "__init__.py"
+        fake.write_text("# no version here\n", encoding="utf-8")
+        monkeypatch.setattr(compatibility, "_VERSION_FILE", fake)
+        assert compatibility.current_core_version() == "3.2.0"
+
+    def test_it_never_raises(self, monkeypatch, tmp_path):
+        # This runs on the install path; an exception here would surface as a
+        # failed update rather than a version mismatch.
+        bad = tmp_path / "__init__.py"
+        bad.write_bytes(b"\xff\xfe\x00 not utf-8 \xff")
+        monkeypatch.setattr(compatibility, "_VERSION_FILE", bad)
+        assert isinstance(compatibility.current_core_version(), str)
+
+
+class TestTheBugItFixes:
+    def test_a_stale_import_no_longer_refuses_a_compatible_plugin(self, monkeypatch, tmp_path):
+        """The exact hardware failure, as a test."""
+        import src
+        monkeypatch.setattr(src, "__version__", "3.2.0")     # what the process holds
+        fake = tmp_path / "__init__.py"
+        fake.write_text('__version__ = "3.3.0"\n', encoding="utf-8")   # what is on disk
+        monkeypatch.setattr(compatibility, "_VERSION_FILE", fake)
+
+        manifest = {"name": "Hockey Scoreboard", "min_ledmatrix_version": "3.3.0"}
+
+        stale_ok, _ = compatibility.check(manifest, src.__version__)
+        assert stale_ok is False, "precondition: the stale value is what refused it"
+
+        fresh_ok, reason = compatibility.check(
+            manifest, compatibility.current_core_version())
+        assert fresh_ok is True, f"the disk version must allow it, got: {reason}"
+
+    def test_it_still_refuses_when_the_core_really_is_too_old(self, monkeypatch, tmp_path):
+        # The gate must not become permissive: a genuinely old core still says no.
+        fake = tmp_path / "__init__.py"
+        fake.write_text('__version__ = "3.2.0"\n', encoding="utf-8")
+        monkeypatch.setattr(compatibility, "_VERSION_FILE", fake)
+        ok, reason = compatibility.check(
+            {"name": "Hockey", "min_ledmatrix_version": "3.3.0"},
+            compatibility.current_core_version())
+        assert ok is False
+        assert "3.2.0" in (reason or "")
+
+
+class TestCallSites:
+    @pytest.mark.parametrize("module", [
+        "src.plugin_system.store_manager",
+        "src.plugin_system.plugin_loader",
+    ])
+    def test_no_gate_binds_the_version_at_import(self, module):
+        """Catch a future call site reintroducing the stale read."""
+        import inspect
+        mod = importlib.import_module(module)
+        source = inspect.getsource(mod)
+        assert "from src import __version__ as core_version" not in source, (
+            f"{module} binds __version__ at import; use "
+            f"compatibility.current_core_version() so a long-lived process "
+            f"sees a core update.")