Support cleaning up broken jobs triggered via Async handlers

[ranjib]

System crash can cause a serf handler to crash in a situation where the state file exist (with process running state) but the process itself is dead. This inhibits all other events (kill, reap, start etc). Provide a way to detect and recover from such scenarios