When we do resubmissions, when a batch of 100 files fails to merge, the missing files are spread across multiple batches if you rerun. Not certain how this can even happen but it would be very useful to keep the original ordering from metacat which is based on the time the files got into the query, perhaps by using a list rather than a map of files?