We are currently able to choose selection rules like "most filled", "most recent", etc when deciding how to pick the winning record during deduplication. However, particularly when using "most filled", it's still possible for there to be duplicates as multiple records could have the same level of "fillness". As far as I can tell, the winner is then chosen at random which could result in different winning records being chosen for the same deduplication group in subsequent runs on the same input files. It would be better if we could add additional rules (e.g. alphabetical) so we could be sure we'd always get the same results from the same input files.