`test_reschedule_concurrent_requests_deadlock` flaky

`test_reschedule_concurrent_requests_deadlock` added in https://github.com/dask/distributed/pull/5379 may be flaky. In https://github.com/dask/distributed/runs/4099410548?check_suite_focus=true (from https://github.com/dask/distributed/pull/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 https://github.com/dask/distributed/blob/8cc428454651caba0e03a99e537f86ac32209395/distributed/tests/test_steal.py#L993-L994 or https://github.com/dask/distributed/blob/8cc428454651caba0e03a99e537f86ac32209395/distributed/tests/test_steal.py#L1013

Perhaps https://github.com/dask/distributed/issues/5370 is not fully closed?

cc @fjetter