What worked for Usenet was contacting the "sysop" of the node the user was posting from, usually upenn, and getting them to have a chat with the user and/or ban them. If their admin wasn't responsive, contact the upstream of their site.

Decentralized policing of users and having one responsible "moderator" per a few hundred users.

Then the internet was opened for commercial use and we got Eternal September and Canter and Siegel.