Submission ID 9dc58cc3-b495-4c61-be5a-28254e180ac8 was started on March 2nd (MutTect2 jobs on 24 REBC pairs) with preemptable=0. We previously commented that these jobs took many hours to actually start. Five of the 24 jobs finished before 7 March 6pm. Between 6pm and 7:40pm EST on 7 March, the remaining 19 jobs must have been restarted by cromwell. These jobs were not "dopplegangers" in that the time stamps in the stdout log are all consistent with being restarted on 7 March and we found no evidence of doggleganer jobs using Eddie's detection tool. Do you expect that cromwell 25 will fix this problem?
Chip & Gordon
@birger @eddieasalinas @gordon123 @chip