+ # queue. If crunch-job exited after losing a race to another
+ # crunch-job process, it exits 75 and we should leave the job
+ # record alone so the winner of the race do its thing.
+ #
+ # There is still an unhandled race condition: If our crunch-job
+ # process is about to lose a race with another crunch-job
+ # process, but crashes before getting to its "exit 75" (for
+ # example, "cannot fork" or "cannot reach API server") then we
+ # will assume incorrectly that it's our process's fault
+ # jobrecord.started_at is non-nil, and mark the job as failed
+ # even though the winner of the race is probably still doing
+ # fine.