Using async process execution in an environment with higher load, we can see that there are longer pauses before the async node gets executed, which makes it seem like the process is getting stuck:

02:46:07,986 ewih DEBUG [org.jbpm.executor.impl.ExecutorImpl] (default task-60) Sending JMS message to trigger job execution for job 14275678
...
02:48:18,219 ewih DEBUG [org.jbpm.executor.impl.jms.JmsAvailableJobsExecutor] (Thread-16 (ActiveMQ-client-global-threads)) received Text message with content 14275678

In this example it takes ~2min to process the JMS message for async execution, but it can even be longer. Thread dumps show that all available KieExecutorMDB threads are busy:

$ grep -c 'KieExecutorMDB.onMessage' *.out
jstack_20210316224029.out:14
jstack_20210316224429.out:15
jstack_20210316224816.out:15

The intent of this JIRA is to make it easier to apply KieExecutorMDB pool size changes in Openshift.