Since the migration might be queued already and be rescheduled when it
fails on a uniqueness error, this should help clearing the background
migration queue faster.