An un-runnable container can get stuck in an infinite loop of lock/unlock cycles (the dispatcher takes it, fails to start it, and the returns it to the queue). It should keep track of the number of times it is locked/unlocked and cancel the container if it exceeds a configured limit.