TRIP - one last shutdown hang #96
Description
I can consistently reproduce a teardown failure (hangs instead) when running the test_cancellation.py::test_nested_multierrors() test with trip but only when trace level logging is enabled via the --ll trace to pytest.
At face value this seems to be a race condition where somehow a couple subactors aren't getting their cancel messages. Monitoring logging and the process list shows this is likely the case. It needs more investigation but I don't think is a show stopper especially given the test is reproducing a case unlikely to be used in practice.