It would be amazing to have a max retry count to pass along with recover_stuck_jobs=true to avoid bad jobs that are looping forever.