Thanks for your reply. The app and jar contains a lot of company internal code which I'm not allowed to share I could try to create a minimal example but I don't know if it will trigger the same behavior. Please let me know if I should do that. After some more debugging (with custom log messages) it must be some race condition. The whole CDI bootstrap happens in threads named "Weld Thread Pool – x" & "MSC service thread x-x". The step "START bootstrap > endInitialization" is then in a thread named "ServerService Thread Pool – xxx" which also fires the initialized event. It seems to work every time when enough log output is produced (which again makes it harder to debug) As as site note: It never failed with WildFly 10.0.1.Final |