Stabilize type-module rebuild test against subscription race

[glen-84]

Summary

• RequestExecutorManagerTests.OnTypesChanged_Should_Not_Grow_CreateTypes_Calls_Exponentially_When_Type_Instance_Registered flaked on CI with a 30 s TaskCanceledException. A TriggerChange() only drives a rebuild while a change-monitor is subscribed, and the Dispose type module change monitor on failed executor creation #9902 fix disposes that monitor when a rebuild fails. The old loop fired each change then waited for the rebuild counter to bump, so under CI contention the next trigger could land after the failed rebuild had torn its monitor down, get dropped, and leave the wait to time out.
• Reworked the test to invert the timing dependency so over-waiting is safe: fire one change while the initial monitor is reliably subscribed, deterministically wait for the failed rebuild to release its subscription (now observable via CountingTypeModule.SubscriberCount), then assert further changes are no-ops. A leaked subscription would instead fan out into additional rebuild attempts.

Test plan

• Ran the test 8× locally on net10.0: passes (~700 ms each).
• Temporarily reintroduced the leak (removed the dispose-on-failure in RequestExecutorManager) and confirmed the test still fails, then restored it, verifying the test remains a regression guard.

[Copilot]

Pull request overview

Stabilizes a flaky RequestExecutorManager type-module rebuild regression test by removing a timing dependency on “change triggers must coincide with an active subscription”, and instead asserting behavior after the failed rebuild has disposed its change-monitor subscription.

Changes:

• Reworked the test flow to trigger one rebuild deterministically, wait for the failed rebuild to release its subscription, then assert subsequent changes are no-ops.
• Updated the test CountingTypeModule to implement ITypeModule directly and expose a SubscriberCount for observing subscription release.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-code-quality]

Code Coverage Overview

Languages: C#

C# / code-coverage/dotnet

The overall coverage in the gai/stabilize-type-m... branch remains at 49%, unchanged from the main branch.

Show a code coverage summary of the most impacted files.

File	main 4691dfa	gai/stabilize-type-m... e197ad3	+/-
/home/runner/wo...sageRecorder.cs	100%	88%	-12%
/home/runner/wo...gerMessage.g.cs	19%	15%	-4%
/home/runner/wo...PromiseCache.cs	81%	78%	-3%
/home/runner/wo...eiveEndpoint.cs	72%	69%	-3%
/home/runner/wo...geDispatcher.cs	77%	75%	-2%
/home/runner/wo...DefaultTopic.cs	90%	89%	-1%
/home/runner/wo...nManagerBase.cs	83%	82%	-1%
/home/runner/wo...qlConnection.cs	89%	92%	+3%
/home/runner/wo...gerMessage.g.cs	15%	20%	+5%
/home/runner/wo...toResetEvent.cs	68%	73%	+5%

---

_{Updated June 17, 2026 15:24 UTC
Code Coverage is in Public Preview. Learn more and provide us with your feedback.}