test_reschedule_concurrent_requests_deadlock flaky

[gjoseph92]

test_reschedule_concurrent_requests_deadlock added in #5379 may be flaky. In https://github.com/dask/distributed/runs/4099410548?check_suite_focus=true (from #5208) it failed on all Windows machines, but no others.

The failures all seem to be gen_cluster timing out after 30s, so I'm not even sure at which step in the test it's deadlocking. I'd guess either

distributed/distributed/tests/test_steal.py

Lines 993 to 994 in 8cc4284

	while not w0.active_keys:
	await asyncio.sleep(0.01)

or

distributed/distributed/tests/test_steal.py

Line 1013 in 8cc4284

await c.gather(futs1)

Perhaps #5370 is not fully closed?

cc @fjetter

[gjoseph92]

This may be particularly flaky on Windows? Once again, it timed out all Windows runs, but no others: https://github.com/dask/distributed/runs/4099410620?check_suite_focus=true

[fjetter]

A more recent failure with test timeout artifact

https://github.com/dask/distributed/runs/4561062302?check_suite_focus=true

windows-latest-3.9-ci1-timeouts.zip