Allow that the renamed blocker thread is terminated

[aeons]

In WorkerThreadNameSpec, if the thread is readded to the compute pool, we still check that it has been renamed and has the correct index, but if the thread has been terminated in the mean time, that is ok as well.

Closes #3704

[armanbilge]

Thanks! Can we actually apply this patch to the 3.5.x branch? Then we'll merge it up into 3.x. Hopefully that should avoid it being clobbered by a merge again...

[aeons]

Changed branch and cherrypicked.

[durban]

It is not clear to me how could the blocker thread exit: the runtimeBlockingExpiration is 1 minute by default. Could this test take such a long time?

[armanbilge]

the runtimeBlockingExpiration is 1 minute by default. Could this test take such a long time?

That's an interesting point. Since we are making a custom IORuntime for this test anyway, what if we just increased that expiration time? I can't remember if the configuration is exposed or hard-coded.

[durban]

Yeah, that's exactly what I was thinking about. It's exposed here: https://github.com/typelevel/cats-effect/blob/series/3.5.x/core/jvm/src/main/scala/cats/effect/unsafe/IORuntimeCompanionPlatform.scala#L57

[aeons]

Switched it to expire threads after 10 minutes. Let's see if this fixes it.

[armanbilge]

Thanks!