The clustered EJB timers seem to be misbehaving in case the timer is used to start new process instances via the startEvent. After talking to swiderski.maciej and ksuta the expected behavior is that the different engine instances (inside different kie-servers) should cope with this, and only one of them should fire the timer. Current behavior is that both (or more) kie-server instances fire the timers.

This issue was observed in kie-server, but it probably is not limited to that.