That still feels comparatively trivial to extract passwords from that scheme. The subsets are small (otherwise users wouldn't be able to respond) so it's simpl

That still feels comparatively trivial to extract passwords from that scheme. The subsets are small (otherwise users wouldn't be able to respond) so it's simple to try all combinations of size, say, 3, until you get the right hash. If the database is compromised, the passwords are then easy to reconstruct.