Stress fixes: framerate/PerGroup timeouts, overlay tick race, DataGrid scroll poll (#397)

Three independent stress-flake fixes (Clusters T, O, C) per
INVESTIGATION.md follow-up to PR #396:

Cluster T — bump FixtureTimeout to 30s for three render-pump-heavy
fixtures whose budgets are tight on loaded CI runners but not
pathological:
  - DataGridParityFixtures.HookPagingFramerateScroll
  - AsyncResourceFramerateFixtures.DataGridEditMutation
  - NativeDockingSmokeFixture.PerGroupDropTargetVisualDemo

Cluster O — fix race in ReconcileHighlightOverlay.RefreshOrAdd.
DispatcherQueueTimer.Stop() can't dequeue a Tick that the dispatcher
queue has already dispatched, so a stale tick can tear down a sprite
the refresh still wants alive. Now each refresh swaps in a fresh timer
with its own Tick lambda; the stale tick checks ReferenceEquals on
ah.Timer and bails when its identity no longer matches.

Cluster C — DataGrid_ScrollPopulatesData was relying on a fixed 800ms
Render(800) for the scroll-settle → fetch → realization chain (not
tracked by _renderPending). Now waits 800ms baseline (preserves the
fetch-trigger window so ScrollPop_MultipleFetches still triggers) then
polls cells for up to 3s.

Validated locally: all 27 OverlayLifecycle_* fixtures pass, all 5
ScrollPop_* sub-checks pass, 15x local stress on both fixtures
together is clean.

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: codemonkeychris

src/Reactor/Hosting/ReconcileHighlightOverlay.cs

@@ -141,8 +141,17 @@ private void RefreshOrAdd(UIElement host, UIElement target, CompositionBrush bru
             existing.Sprite.Opacity = opacity;
             existing.Sprite.Size = size;
             existing.Sprite.Offset = offset;
+            // Dispatcher-queue contract: DispatcherQueueTimer.Stop() does NOT
+            // dequeue a Tick that was already enqueued before the call. If the
+            // original interval elapsed concurrently with this refresh, the old
+            // Tick can fire AFTER Start() and tear down a sprite that the user
+            // still expects to see. Swap in a fresh timer with its own Tick
+            // lambda so the old tick — if it fires — finds the active entry
+            // no longer owns it, and bails (see the identity check below).
             try { existing.Timer.Stop(); } catch { }
-            try { existing.Timer.Start(); } catch { }
+            var refreshedTimer = CreateExpiryTimer(target, existing);
+            existing.Timer = refreshedTimer;
+            refreshedTimer.Start();
             return;
         }
 
@@ -157,17 +166,31 @@ private void RefreshOrAdd(UIElement host, UIElement target, CompositionBrush bru
         sprite.Brush = brush;
         _container.Children.InsertAtTop(sprite);
 
+        var entry = new ActiveHighlight { Sprite = sprite };
+        var timer = CreateExpiryTimer(target, entry);
+        entry.Timer = timer;
+        _active[target] = entry;
+        timer.Start();
+        newBudget--;
+    }
+
+    private DispatcherQueueTimer CreateExpiryTimer(UIElement capturedTarget, ActiveHighlight owner)
+    {
         var timer = _dispatcherQueue.CreateTimer();
         timer.Interval = TimeSpan.FromMilliseconds(_holdDurationMs);
         timer.IsRepeating = false;
 
-        var capturedTarget = target;
         var container = _container;
         timer.Tick += (s, _) =>
         {
             try
             {
-                if (_active.TryGetValue(capturedTarget, out var ah))
+                // Identity guard: a Refresh between this tick's enqueue and its
+                // dispatch will have swapped in a new timer; the stale tick must
+                // not tear down the sprite that the new timer still owns.
+                if (_active.TryGetValue(capturedTarget, out var ah)
+                    && ReferenceEquals(ah, owner)
+                    && ReferenceEquals(ah.Timer, s))
                 {
                     try { container.Children.Remove(ah.Sprite); } catch { }
                     try { ah.Sprite.Dispose(); } catch { }
@@ -179,10 +202,7 @@ private void RefreshOrAdd(UIElement host, UIElement target, CompositionBrush bru
                 try { ((DispatcherQueueTimer)s).Stop(); } catch { }
             }
         };
-
-        _active[target] = new ActiveHighlight { Sprite = sprite, Timer = timer };
-        timer.Start();
-        newBudget--;
+        return timer;
     }
 
     // ─────────────────────────────────────────────────────────────────────

tests/Reactor.AppTests.Host/SelfTest/Fixtures/AsyncResourceFramerateFixtures.cs

@@ -417,6 +417,10 @@ public override Element Render()
 
     internal class DataGridEditMutation(Harness h) : SelfTestFixtureBase(h)
     {
+        // 60-frame mutation pump + 30-frame drain; same wall-clock-floor reasoning
+        // as HookPagingFramerateScroll. See INVESTIGATION.md Cluster T2.
+        public override TimeSpan FixtureTimeout => TimeSpan.FromSeconds(30);
+
         public override async Task RunAsync()
         {
             await Task.Run(() =>

tests/Reactor.AppTests.Host/SelfTest/Fixtures/DataGridParityFixtures.cs

@@ -262,6 +262,11 @@ private async Task RunInner()
     /// </summary>
     internal class HookPagingFramerateScroll(Harness h) : SelfTestFixtureBase(h)
     {
+        // 60-frame programmatic scroll has a ~3.4 s mandatory Render(ms) wall-clock
+        // floor; loaded CI runners slow per-frame work 2–4x and trip the default 15 s
+        // budget without anything being wedged. See INVESTIGATION.md Cluster T2.
+        public override TimeSpan FixtureTimeout => TimeSpan.FromSeconds(30);
+
         public override Task RunAsync() => WithHookPaging(true, RunInner);
 
         private async Task RunInner()

tests/Reactor.AppTests.Host/SelfTest/Fixtures/DataGridScrollFixtures.cs

@@ -113,25 +113,35 @@ public override async Task RunAsync()
             H.Check("ScrollPop_ScrollViewerFound", sv is not null);
             if (sv is null) return;
 
-            // 2. Scroll to a position deep in the list (row ~200)
+            // 2. Scroll to a position deep in the list (row ~200).
+            // Baseline wait (preserves the scroll-settle window so the
+            // ScrollPop_MultipleFetches sanity assertion still triggers fetches),
+            // then poll cells in case the realization runs long. The original
+            // fixed 800ms Render flaked as `cells=0` under CI load — see
+            // INVESTIGATION.md Cluster C.
             sv.ChangeView(null, 7200, null, disableAnimation: true);
-            // Wait for: scroll event → EnsureRangeLoaded → async fetch → settle timer → render
             await Harness.Render(800);
-
-            // 3. The key assertion: rows at the scroll target should have real data,
-            //    not placeholders. Look for "Emp-" prefix in visible TextBlocks.
-            var visibleEmpCells = H.FindAllControls<TextBlock>(
-                tb => tb.Text?.StartsWith("Emp-") == true);
+            var visibleEmpCells = H.FindAllControls<TextBlock>(tb => tb.Text?.StartsWith("Emp-") == true);
+            var deadline1 = Environment.TickCount64 + 3_000;
+            while (visibleEmpCells.Count < 4 && Environment.TickCount64 < deadline1)
+            {
+                await Harness.Render(100);
+                visibleEmpCells = H.FindAllControls<TextBlock>(tb => tb.Text?.StartsWith("Emp-") == true);
+            }
 
             H.Check($"ScrollPop_DataVisible (cells={visibleEmpCells.Count})",
                 visibleEmpCells.Count >= 4);
 
             // 4. Scroll to a completely different position
             sv.ChangeView(null, 14400, null, disableAnimation: true);
             await Harness.Render(800);
-
-            var visibleEmpCells2 = H.FindAllControls<TextBlock>(
-                tb => tb.Text?.StartsWith("Emp-") == true);
+            var visibleEmpCells2 = H.FindAllControls<TextBlock>(tb => tb.Text?.StartsWith("Emp-") == true);
+            var deadline2 = Environment.TickCount64 + 3_000;
+            while (visibleEmpCells2.Count < 4 && Environment.TickCount64 < deadline2)
+            {
+                await Harness.Render(100);
+                visibleEmpCells2 = H.FindAllControls<TextBlock>(tb => tb.Text?.StartsWith("Emp-") == true);
+            }
 
             H.Check($"ScrollPop_DataVisibleAfterSecondScroll (cells={visibleEmpCells2.Count})",
                 visibleEmpCells2.Count >= 4);

tests/Reactor.AppTests.Host/SelfTest/Fixtures/NativeDockingSmokeFixture.cs

@@ -1002,6 +1002,11 @@ public override async Task RunAsync()
     /// </summary>
     internal class PerGroupDropTargetVisualDemo(Harness h) : SelfTestFixtureBase(h)
     {
+        // 20 mount cycles × ~700 ms each on loaded CI runners can overshoot the
+        // default 15 s budget; the demo is not pathological, just paced for human
+        // visibility. See INVESTIGATION.md Cluster T1.PerGroup.
+        public override TimeSpan FixtureTimeout => TimeSpan.FromSeconds(30);
+
         public override async Task RunAsync()
         {
             var host = H.CreateHost();