#11771 make cfreactor less flaky

[glyph]

Scope and purpose

Fixes #11771

This is the change that appeared to have made a difference isolated and extracted from #12150 .

[glyph]

Both of the failures are in twisted.internet.test.test_tcp.AbortConnectionTests_CFReactorTests.test_fullWriteBuffer, which also fails sometimes in the asyncioreactor too: #12151

so I think this fix actually resolves the issue

[glyph]

please review

[adiroiban]

Thanks. Many, many thanks for working on this.

Were you able to reproduce the errors on your local system?

I was trying to reproduce them on a macmini m1, without luck.

---

I think that we should just merge it and see if works.

[adiroiban]

Previous comment was an "approve".

[glyph]

Were you able to reproduce the errors on your local system?

I was not. But in every other case I managed to reproduce the error in CI after a minimum of 5 tries.

The race condition here is extremely subtle, and I couldn't even think of a white-box test to reproduce it, but basically if something were to call _scheduleSimulate without force after calling it with force, so logically I think I see how this hang could happen.