quartz scheduler can die

If an Error is thrown in any code run by the quartz scheduling system, the thread executing the code silently dies, and is not restarted. after this happens N times, where N is the number of threads started by quartz (4, by default) the scheduler stops and no further scheduled activities are run. For example, mail is no longer delivered, but just piles up in the queue.

For some Errors (OutOfMemory, say) this might make a sort of sense. The ship is sinking anyhow, so there's no sense in doing much about it. However, other errors (NoClassDefFound, for example) should not, under any circumstance, cause the system to stop running.

For example, JIRA-2356 describes how a totally unmodified JIRA installation will throws Errors in the "Backup Service", leading eventually to the state described here.