Fix flaky TestResourceManagerIsSafeForConcurrentAccessAndEnumeration timeout

[lewing]

Problem

TestResourceManagerIsSafeForConcurrentAccessAndEnumeration is flaky (#125448, 3 hits in 7 days). This same test has been filed as a Known Build Error three times over 3+ years (#80277, #86013, #125448).

The test spins up 10 threads that concurrently enumerate a resource set with Thread.Sleep(1) between entries, then waits with:

Assert.True(Task.WaitAll(tasks, TimeSpan.FromSeconds(30)));

On loaded CI agents, 30 seconds isn't always enough. The failure message is unhelpful — just Assert.True() Failure: Expected: True, Actual: False — because Task.WaitAll swallows any actual thread-safety exceptions.

Fix

Task allTasks = Task.WhenAll(tasks);
Task completed = await Task.WhenAny(allTasks, Task.Delay(TimeSpan.FromSeconds(120)));
Assert.True(completed == allTasks, "Timed out waiting for concurrent enumeration tasks");
await allTasks; // propagates any real exceptions

Three improvements:

1. Task.WhenAll instead of Task.WaitAll — if a real thread-safety bug is hit, the actual exception propagates instead of being swallowed behind Assert.True(false)
2. 120s instead of 30s — 4x more headroom for loaded CI agents while still providing a timeout safety net
3. Clear timeout message — on timeout, the assertion message says what happened instead of a bare Assert.True failure

The WhenAny/Task.Delay pattern is used instead of Task.WhenAll(...).WaitAsync(...) for .NET Framework TFM compatibility. The method signature changes from void to async Task, which xUnit handles natively.

Fixes #125448

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-resources
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

This PR aims to de-flake PreserializedResourceWriterTests.TestResourceManagerIsSafeForConcurrentAccessAndEnumeration by improving how the test waits for concurrent worker tasks to complete, increasing timeout headroom and (intended to) improve exception surfacing during failures.

Changes:

• Converted the test from void to async Task.
• Replaced Task.WaitAll(..., 30s) with await Task.WhenAll(...).WaitAsync(120s).

---

You can also share your feedback on Copilot code review. Take the survey.

[lewing]

Fixed — replaced Task.WhenAll(...).WaitAsync(120s) with Task.WhenAny(Task.WhenAll(tasks), Task.Delay(120s)) pattern that compiles on all TFMs including $(NetFrameworkCurrent). Thanks for catching that!

[Copilot]

Pull request overview

Reduces flakiness in TestResourceManagerIsSafeForConcurrentAccessAndEnumeration by increasing the timeout and improving exception propagation when concurrent enumeration fails.

Changes:

• Makes the test method async Task to enable async waiting patterns in xUnit.
• Replaces Task.WaitAll(..., 30s) with an async timeout-based wait around Task.WhenAll(...).
• Adds a clearer assertion message on timeout and awaits the aggregated task to propagate real exceptions.

---

You can also share your feedback on Copilot code review. Take the survey.

[danmoseley]

/ba-g generic chrome crash.

[danmoseley]

merging -- i did /ba-g and it still won't pass 5 mins later.

[danmoseley]

ah, I apparently don't have the power 😐

[danmoseley]

/ba-g chrome generic crash.