It's not really "nepotic". The goal was to reduce the number of accounts you need to look at to some manageable number. Ideally you discover some root spammer account and ban the entire tree under the assumption they're all spam accounts. Then you ban the guy who was dumb enough to invite a spammer into the community since they clearly cannot be trusted. If you can't or won't ban them, you take away their invitation privileges.

If the problematic account is too high up in the tree, then they probably have way too many descendants which increases the false positive rate which means you can't conclude that they're bad based on the fact they're invitees of a bad account. Essentially, the spammers diluted themselves inside a huge number of maybe good users so that you can't blanket ban them without collateral damage. In that case you're better off banning the bad account individually and looking for new bad roots further down the tree.