It's a nightmare for harassment though, treating all blocking as a sissy snowflake shield renders it essentially useless for circumstances where the other user is really malicious.

In that case having a system like Mastodon's that locks them out of Replying is better than simply covering them up. We can't really rely on Instance admins to block all malicious users or defederate from all malicious platforms so there needs to be a certain level of protection from the user's side.

I do think that Blocking and Restricting (preventing a user from interacting) should be different functions, you block communities and users because you don't want to see them, but you restrict users because they are hostile, malicious, or make bad-faith arguments.