Smallest first (default) returns the smallest numbers with at least the given digits — the frontier to work through systematically. Random sample returns numbers of exactly that many digits from a random point, so multiple contributors don't all pull the same rows. Either way it's one bounded index scan — light on the server. The batch is capped at 50000 numbers and ~10 MB of text (fewer rows are returned for very large sizes).

Found a factor or a certificate? Submit it back via Report.