perf(capture): slice 8 — buffer ring + paste-cache Epoch; gen2 visibility (take-11 spikes)

Take 11 (c10ce06c) validated the off-UI architecture: typical frames land
work ~10ms + wait ~6.8ms = 16.7 exactly on the deadline; 212/300 best yet.
The ENTIRE remaining gap is periodic 35-65ms render spikes that WORSENED
across the take (189 -> 147) — the signature of gen2 GC pauses. Biggest
churner is structural: the screen capture minted a fresh ~8.3MB byte[] per
DWM frame (~500MB/s of LOH), a producer OBS never does (it owns fixed
surface pools).

- ScreenCaptureFrameSource: 4-deep buffer ring with size-matched slots (a
  <=17ms consumer cannot be lapped at 60Hz) + reused downscale row scratch.
- VideoFrame.Epoch: monotonic per producer frame. The paste cache keys on
  array IDENTITY, so recycled arrays MUST be distinguished — epoch joins the
  PasteKey. Producers handing fresh arrays leave it 0 (key unchanged effect).
- Stats print 'gen2 +N' per 5s window: next take acquits or convicts GC
  without another guess (rule: prove the stage).
- Test (the ONE): PasteCache_RecycledArrayWithNewEpoch_ReRasterizes_NotStaleHits
  — same array, new content, bumped epoch; fails on the old key by
  construction. 37/37 compositor/pump, clean build.
- Next suspect if gen2 stays hot: the 10Hz WebView2 capture (full-canvas PNG
  decode + fresh arrays on the UI thread) — recorded, untouched.

Creator audio ask queued in the same working session (+40% post-mix master
gain before the -1dBFS limiter) lands as its own commit next.

Services/Compositor/SceneCompositor.cs

@@ -19,10 +19,11 @@ public sealed class SceneCompositor
     // chat/web/image layers are identical frame after frame. Same shape as OBS (the
     // source surface is cached, the compositor blits): a scaled/round-clipped/mirrored
     // layer rasterizes once and is pasted (row-copy/blend) on every later tick.
-    // Keyed by the SOURCE FRAME IDENTITY — every producer (capture, webcam, chat
-    // renderer, web capture, static cache) hands out immutable byte[], so "same pixels
-    // array + same target rect" ⇒ same raster. Bounds-checked eviction clears wholesale.
-    private sealed record PasteKey(byte[] Pixels, int SrcW, int SrcH, int DstW, int DstH, bool Round, bool Mirror);
+    // Keyed by the SOURCE FRAME IDENTITY (array + Epoch) — producers hand out fresh
+    // immutable byte[], or recycled-ring arrays whose monotonic Epoch distinguishes
+    // each generation (screen capture, take-11 fix), so "same array+epoch + same
+    // target rect" ⇒ same raster. Bounds-checked eviction clears wholesale.
+    private sealed record PasteKey(byte[] Pixels, long Epoch, int SrcW, int SrcH, int DstW, int DstH, bool Round, bool Mirror);
     private static readonly Dictionary<PasteKey, VideoFrame> _pasteCache = new();
     private static readonly object _pasteGate = new();
     private const int MaxPasteEntries = 48;
@@ -284,7 +285,7 @@ public sealed class SceneCompositor
         VideoFrame? raster = null;
         lock (_pasteGate)
         {
-            key = new PasteKey(frame.BgraPixels, frame.Width, frame.Height, dw, dh, isRound, mirror);
+            key = new PasteKey(frame.BgraPixels, frame.Epoch, frame.Width, frame.Height, dw, dh, isRound, mirror);
             _pasteCache.TryGetValue(key, out raster);
         }
 

Services/ScreenCaptureFrameSource.cs

@@ -30,6 +30,18 @@ public sealed class ScreenCaptureFrameSource : IScreenCaptureSource
     private bool _framePending;
     private DateTime _lastErrorLog = DateTime.MinValue;
 
+    // Buffer recycling (take-11 spikes, 2026-09-04): a fresh ~8.3MB byte[] per DWM
+    // frame ≈ 500MB/s of LOH churn — the gen2 pauses it forces surfaced as the
+    // "worst render 35-65ms" spikes that capped fps at ~42 long after the compositor
+    // itself was fast. A 4-deep ring rotated round-robin is never lapped by a
+    // ≤17ms consumer at 60Hz; each hand-out carries an Epoch so identity-keyed
+    // consumers (the compositor's paste cache) cannot false-hit a recycled array.
+    private readonly byte[]?[] _frameRing = new byte[4][];
+    private int _ringNext;
+    private long _epoch;
+    private byte[]? _row0;
+    private byte[]? _row1;
+
     // The composition master frame (see ai.md "Resolution tiers"): the background
     // is an input layer, so we never hold a CPU frame bigger than the master.
     private const int MaxBackgroundWidth = 1920;
@@ -157,7 +169,26 @@ public sealed class ScreenCaptureFrameSource : IScreenCaptureSource
         }
     }
 
-    private static VideoFrame CopyToVideoFrame(SoftwareBitmap bitmap)
+    private byte[] RentRingBuffer(int size)
+    {
+        for (var tries = 0; tries < _frameRing.Length; tries++)
+        {
+            var idx = (_ringNext + tries) % _frameRing.Length;
+            var buf = _frameRing[idx];
+            if (buf is { Length: var len } && len == size)
+            {
+                _ringNext = (idx + 1) % _frameRing.Length;
+                return buf;
+            }
+        }
+        var slot = _ringNext;
+        _ringNext = (slot + 1) % _frameRing.Length;
+        var fresh = new byte[size];
+        _frameRing[slot] = fresh;
+        return fresh;
+    }
+
+    private VideoFrame CopyToVideoFrame(SoftwareBitmap bitmap)
     {
         var sw = bitmap.PixelWidth;
         var sh = bitmap.PixelHeight;
@@ -178,21 +209,23 @@ public sealed class ScreenCaptureFrameSource : IScreenCaptureSource
             var dw = Math.Max(1, (int)(sw * scale));
             var dh = Math.Max(1, (int)(sh * scale));
             // DWM delivers an opaque surface (alpha 255); bilinear keeps it 255.
-            return new VideoFrame(dw, dh, DownscaleBgra(data, sw, sh, srcStride, dw, dh)) { IsOpaque = true };
+            var scaled = RentRingBuffer(dw * dh * 4);
+            return new VideoFrame(dw, dh, DownscaleBgra(data, sw, sh, srcStride, dw, dh, scaled))
+                { IsOpaque = true, Epoch = ++_epoch };
         }
 
-        var pixels = new byte[count];
+        var pixels = RentRingBuffer(count);
         Marshal.Copy(data, pixels, 0, pixels.Length);
-        return new VideoFrame(sw, sh, pixels) { IsOpaque = true };
+        return new VideoFrame(sw, sh, pixels) { IsOpaque = true, Epoch = ++_epoch };
     }
 
     // Bilinear downscale to the master frame. Reads each source row pair through
     // Marshal.Copy (no unsafe), writing tightly packed BGRA output.
-    private static byte[] DownscaleBgra(IntPtr src, int sw, int sh, int srcStride, int dw, int dh)
+    private byte[] DownscaleBgra(IntPtr src, int sw, int sh, int srcStride, int dw, int dh, byte[] dst)
     {
-        var row0 = new byte[srcStride];
-        var row1 = new byte[srcStride];
-        var dst = new byte[dw * dh * 4];
+        // Row scratch is per-capture-thread and reused across frames (same churn lesson).
+        var row0 = _row0 != null && _row0.Length >= srcStride ? _row0 : (_row0 = new byte[srcStride]);
+        var row1 = _row1 != null && _row1.Length >= srcStride ? _row1 : (_row1 = new byte[srcStride]);
         var xs = sw / (double)dw;
         var ys = sh / (double)dh;
 

Services/VideoFrame.cs

@@ -13,6 +13,13 @@ public sealed class VideoFrame
     public byte[] BgraPixels { get; }
     public int Stride => Width * 4;
 
+    /// <summary>Producer frame epoch — a monotonically increasing number stamped by
+    /// sources that RECYCLE their pixel buffers (the screen-capture ring, take-11 spike
+    /// fix). Consumers that key caches on buffer identity (the compositor's paste cache)
+    /// MUST include it, or a recycled array false-hits with stale content. Producers
+    /// that hand out fresh arrays per frame leave it 0 — identity alone is then enough.</summary>
+    public long Epoch { get; init; }
+
     /// <summary>Producer contract: every alpha byte in <see cref="BgraPixels"/> is 255
     /// (DWM capture surfaces and MediaCapture video carry no alpha — the OS fills 255).
     /// Lets the compositor take a straight-copy fast path for a full-canvas opaque