This issue is caused by wildfly-core#PR6117 and WFCORE-6894. I can't confirm that PR18098 could fix the issue, because there are conflicts in this PR. Anyway, this is a regression, so I set the priority to blocker.

Scenario where server hangs:

• Before the server is commanded to graceful shutdown, a new transaction is started. Server should not shutdown until tx is finished or timeout is gone.
• Procedure:
  • start transaction
  • 1. remote ejb call updates database and commands server to gracefully shutdown with 30s timeout
  • 2. remote ejb call is part of the same tx so it should proceed
  • commit tx
  • check that both ejb calls were successful

Steps to reproduce:
see my first comment on this jira