wondering what 'goodness' metrics are for replication. one is "how many jobs were replicated but the first submission executed (so the replication was in some sense wasted)" I'd be interested in ideas for other metrics --