What I'm advocating for is implementing communities/groups at a client level, which is largely what is happening today. Presumably clients will add actual moderation functionality, giving admins the ability to stop users from posting in these communities, remove posts, etc.

This is fine. The difference is that when a user gets banned from such a community, it doesn't affect their followers' ability to see their posts at all. They can also start their own community for the same topic if they so choose with a couple of clicks.

However, if community relays become a thing and we're presumably only subscribed to relays that we have an interest for (since otherwise the whole idea doesn't work), then getting banned from such a relay will also impact many user's ability to see your posts. It also vastly restricts your ability to start your own community, since presumably you need to have your own relay to do it, and you need to find some way of reaching a bunch of users who are isolated on network islands.