If Artemis runs out of threads with the JVM signalling this with an OOME error, the process continues in a broken state logging exceptions. It would be better if the process exited so the running environment (Kubernetes, job control etc) can restart it.

I read from JDK-8155004[1] that JVM flag XxxxOnOutOfMemoryError is not applied to this sort of OOME.

See https://issues.apache.org/jira/browse/ARTEMIS-2623