Wait for in-flight health check publishers in HealthCheckPublisherHostedService.StopAsync

[SergioAlmeida29]

Fixes #67304

Problem

HealthCheckPublisherHostedService.StopAsync disposes its timers but doesn't wait for a health-check publish that is already running, so IHealthCheckPublisher.PublishAsync can keep running after the host has stopped.

The timer callback ran RunAsync as fire-and-forget — an async void lambda bound to TimerCallback — so nothing tracked the task. Timer.Dispose() (and even Dispose(WaitHandle) / DisposeAsync()) only wait for the synchronous portion of the callback, which ends at the first await.

Change

• Track the Task returned by each timer-driven RunAsync in a set.
• StopAsync awaits the in-flight tasks after disposing the timers, bounded by its CancellationToken (the host links this to ShutdownTimeout). Mirrors BackgroundService.StopAsync: #if NET uses Task.WaitAsync(token) with ConfigureAwaitOptions.SuppressThrowing; older TFMs use Task.WhenAny + token.Register.
• Publishers that ignore the token are still not force-stopped; overlapping-execution and scheduling behavior is unchanged.

Behavior note

StopAsync may now block up to the host shutdown timeout while a publish drains. This matches the IHostedService contract and BackgroundService, and is not an API or binary breaking change.

Tests

Two regression tests driven by the real timer. StopAsync_WaitsForInFlightPublisher_BeforeReturning fails without the fix (StopAsync completes while the publish is still parked).

cc @danroth27 @BrennanConroy

[dotnet-policy-service]

Thanks for your PR, @SergioAlmeida29. Someone from the team will get assigned to your PR shortly and we'll get it reviewed.

[Copilot]

🟡 Changes recommended

There is a shutdown race where a timer callback can start a new tracked RunAsync after StopAsync snapshots an empty set and returns, allowing publish work to outlive shutdown.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR updates HealthCheckPublisherHostedService.StopAsync to await timer-triggered health check publishing work that’s already in flight during shutdown, preventing publishers from continuing after the host has stopped.

Changes:

• Track each timer-driven RunAsync invocation in a shared set and await remaining tasks during StopAsync, bounded by the caller’s cancellation token/shutdown timeout.
• Replace the timer’s async void callback pattern with a fire-and-forget Task method that records/removes the in-flight operation.
• Add regression tests validating StopAsync waits for an in-flight publisher, and that a canceled stop token does not wait.

File summaries

File	Description
src/HealthChecks/HealthChecks/src/HealthCheckPublisherHostedService.cs	Track timer-triggered RunAsync tasks and await in-flight publishes during StopAsync.
src/HealthChecks/HealthChecks/test/HealthCheckPublisherHostedServiceTest.cs	Add regression tests covering shutdown waiting and cancellation-bounded stop behavior.

Review details

• Files reviewed: 2/2 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[gfoidl]

Suggested change

	private readonly object _runningTasksLock = new object();
	private readonly Lock _runningTasksLock = new();

or does this build for older targets w/o Lock-type?