Don't update the job database time stamp on job purge. It was triggering

state saves, backfill scheduler runs, etc. when there was not a significant
state change that should trigger those events.