feat(vegas): prefetch_gate on by default, after an A/B/C soak on hdpi

Two runs per arm, about 81,000 frames each, order A B C C B A:

  A  step 1 as is            0.90% late, 20.1 per 10k two+ refreshes late
  B  switch_interval_ms 1    0.78% late, 15.8 per 10k
  C  prefetch_gate           0.60% late,  2.5 per 10k

No freezes in any arm, and the next group was ready at every strip
extension, so parking the prefetch thread (3-6s per 8-minute run) cost
nothing visible. The gate is now on unless vegas_scroll.prefetch_gate is
false; on a stock binding it cannot work and says so at INFO once a run
rather than warning on every install. switch_interval_ms stays off.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/vegas_mode/config.py

@@ -87,16 +87,17 @@ class VegasModeConfig:
     # 5ms. Plugin rendering on the prefetch thread and plugin updates hold the
     # GIL in Pillow and Python code, and a frame waiting its turn for 5ms at a
     # time misses its refresh. 0 leaves the interpreter default alone.
-    # Experimental: measured with scripts/frame_soak.py before it gets a default.
+    # Experimental. On hdpi it did less than prefetch_gate (0.90% -> 0.78% late
+    # against 0.60%; see docs/OFFSCREEN_RENDERING.md), so it stays off.
     switch_interval_ms: float = 0.0
 
     # Let the prefetch thread run Python only while the render thread is
     # blocked waiting for vsync, and park it the rest of the time, so the
     # render thread never waits for the GIL when its refresh comes round. Needs
-    # a binding that releases the GIL in SwapOnVSync; ignored otherwise.
-    # Experimental: see src/common/render_gate.py and measure with
-    # scripts/frame_soak.py before it gets a default.
-    prefetch_gate: bool = False
+    # a binding that releases the GIL in SwapOnVSync; off otherwise. On hdpi
+    # it cut frames two or more refreshes late eightfold, and late frames
+    # overall from 0.90% to 0.60%. See src/common/render_gate.py.
+    prefetch_gate: bool = True
 
     # Keep one continuous strip, extending it with the next group of plugins as
     # the scroll approaches the end, instead of composing a fresh strip and
@@ -229,7 +230,7 @@ class VegasModeConfig:
             continuous_scroll=vegas_config.get('continuous_scroll', True),
             offscreen_prefetch=bool(vegas_config.get('offscreen_prefetch', True)),
             switch_interval_ms=float(vegas_config.get('switch_interval_ms', 0.0) or 0.0),
-            prefetch_gate=bool(vegas_config.get('prefetch_gate', False)),
+            prefetch_gate=bool(vegas_config.get('prefetch_gate', True)),
             extend_threshold_screens=float(
                 vegas_config.get('extend_threshold_screens', 2.0)),
             auto_trim=vegas_config.get('auto_trim', True),

src/vegas_mode/coordinator.py

@@ -340,12 +340,14 @@ class VegasModeCoordinator:
         if getattr(self.display_manager, 'render_gate', None) is not None:
             return
         releases = render_gate.swap_releases_gil()
+        if releases is None:
+            logger.debug("Vegas: no prefetch gate -- no hardware binding loaded")
+            return
         if not releases:
-            logger.warning(
-                "Vegas: prefetch_gate ignored -- %s",
-                "this rgbmatrix binding keeps the GIL in SwapOnVSync "
-                "(scripts/build_rgbmatrix_nogil.sh)" if releases is False
-                else "no hardware binding loaded")
+            # On by default, so this is every stock install: say so once per
+            # run, not as a warning.
+            logger.info("Vegas: no prefetch gate -- this rgbmatrix binding keeps "
+                        "the GIL in SwapOnVSync (scripts/build_rgbmatrix_nogil.sh)")
             return
         gate = render_gate.RenderGate()
         # Locks the render thread takes too: never park the prefetch holding one.