Condor jobs starting several times after, apparently, having completed
Related SNOW incident: INC000000544058
This is a special rare case since, each time the jobs were started, they had the time to tell the sam projects they had completed so it looks like the followed regular ending path.
Ken and Kevin explanation: It must have been 'network problems' because it all happened in a matter of 2 minutes (a job running in different nodes four times).
Robert Illingworth: To know better, someone should check condor internal logs.
Dennis Box: Joe Boyd would be right person to ask about this.