hangs/failures in fail test cases
The new fail tests in the SSG repo (
ssg-test-fail-pmix) either fail or hang sporadically, at least in my testing on Cooley using PMIx for process launch.
In case of failures, there is no error output or core dump. The failures only occur at some point after which the process is forced to fail (either by explicitly destroying the group or by having PMIx inject and respond to the failure). This warrants further investigation, as it seems likely we would see similar behavior even when processes are explicitly asking to leave a group.