[Nacl-HCl]

Fixes #3275

The issue or feature being addressed

#3275

Details on the issue fix or feature implementation

ScheduledTaskExecutor starts its background processing loop with
Task.Run(StartProcessingAsync) in its constructor, which flows the ambient
ExecutionContext into the loop. The loop waits on its semaphore for as long
as the pipeline lives, so every AsyncLocal value of the code that happened
to build the pipeline stays reachable for the pipeline's lifetime (e.g. the
first user's circuit in a Blazor Server app is never collected after it is
disposed).

The loop is infrastructure that belongs to no caller, and the scheduled tasks
are the breaker's own state-change callbacks, which a caller's ambient
AsyncLocal values should not reach either. The constructor now starts the
loop with the flow suppressed:

using (ExecutionContext.SuppressFlow())
{
    ProcessingTask = Task.Run(StartProcessingAsync);
}

Confirm the following

• I started this PR by branching from the head of the default branch
• I have targeted the PR to merge into the default branch
• I have included unit tests for the issue/feature
• I have successfully run a local build
• I have the experience and knowledge necessary to understand, review, and validate all content in this PR.
• This PR has content that I did not fully write myself.
  • I used AI to assist in writing or reviewing the content of this PR.

Verification

• New regression test Constructor_DoesNotCaptureAmbientExecutionContext
  (in ScheduledTaskExecutorTests): builds an executor while an
  AsyncLocal<object?> holds a value, then asserts the value is collectable
  while the executor (and its background task) is kept alive.
  • before the fix: fails (IsAlive is True — the leak reproduces)
  • after the fix: passes (5/5 runs, no flakiness)
• Polly.Core.Tests full suite (net10.0): 782/782 passed, 0 failed
• Polly.Core builds clean for all target frameworks
  (net8.0;net6.0;netstandard2.0;net472;net462), 0 warnings, 0 errors

Note

This PR description was prepared with AI assistance (Muse). The fix and the
regression test were verified by running the reproduction and the test suite
locally as described above.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.02%. Comparing base (d065fd7) to head (b07d0ff).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3284   +/-   ##
=======================================
  Coverage   98.02%   98.02%           
=======================================
  Files         321      321           
  Lines        7099     7103    +4     
  Branches     1089     1089           
=======================================
+ Hits         6959     6963    +4     
  Misses         98       98           
  Partials       42       42           

Flag	Coverage Δ
linux	97.97% <100.00%> (+<0.01%)	⬆️
macos	98.01% <100.00%> (+<0.01%)	⬆️
windows	98.02% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

[github-actions]

Thanks for your contribution @Nacl-HCl - the changes from this pull request have been published as part of version 8.8.1 📦, which is now available from NuGet.org 🚀