If any federated banning networks do pop up, I'd expect them to form groups, with different groups having different standards. And the idea being that if someone's banned from one place with similar standards, the rest of the group probably wouldn't welcome the content.

It'll come down to places and groups being reasonable, and not banning for stupid reasons (at least by that group's standards). And if they are unreasonable, it'll reflect on the group, as nobody would bother posting to those instances any more.

And in a way, the ultimate "ban" will be with the host instance, similarly to email.
An admin at lemmy.world might get a report that an account is spreading csam links everywhere, and to consider banning them, for example.