1) introduced a global configuration variable job.cancel.threshold.minutes
2) Default value for the job.cancel.threshold.minutes is 60 minutes, if a job has been staying in processing state for over 60 minutes, system will forcely remove it from the queue to allow others to run