Uses memoized min/max for user id. Automatic batch size will halve
smartly depending on the particular gitlab instances data