In fact, trust isn’t manually handled, it’s handled based on how similarly you act vs others (i.e. you both upvote/downvote similarly, flag posts similarly, etc), and I’m deciding whether making this based on community makes sense (i.e. you trust user A on community X, but not on community Y).

Wouldn’t this just create an impenetrable filter bubble/echo chamber where you see nothing else than content you 100% agree with?