Fix test_worker_waits_for_scheduler

[mrocklin]

Remove asyncio.wait_for from test.
It seems to be malfunctioning very rarely.
We replace it with a less sophisticated task + sleep + cancel.

Fixes #6153

[fjetter]

This has a lower level root cause since we're using asyncio.wait_for internally in a bunch of places, particularly the core.connect. The way asyncio.wait_for handles cancellation is subject to a race condition.

I'm fine with rewriting the test if that resolves things in this specific case

[mrocklin]

This has a lower level root cause since we're using asyncio.wait_for internally in a bunch of places, particularly the core.connect. The way asyncio.wait_for handles cancellation is subject to a race condition.

Is this python/cpython#86296 ?

[mrocklin]

Or, more broadly, is there an issue somewhere for this? (doing a search on github for wait_for is less helpful than you might expect)

[fjetter]

Sorry, I meant to copy the link... #6064

Is this python/cpython#86296 ?

This might be related but I need to read it a bit more thoroughly...

[mrocklin]

Heh, this example is also a good case study in how you and I approach issues. I find a fix for the immedate pain in a few minutes, but ignore the deeper problem. You address the deeper problem, but things take a while. It's probably good that we're both here 🙂

[github-actions]

Unit Test Results

       16 files  ±0         16 suites  ±0   7h 15m 48s ⏱️ - 23m 5s
  2 745 tests ±0    2 665 ✔️ +4       80 💤 ±0  0 ❌  - 4 
21 845 runs  ±0  20 810 ✔️ +5  1 035 💤 ±0  0 ❌  - 5 

Results for commit 4d4b318. ± Comparison against base commit fc5b460.