@anthropy @ProfWolf
I think the solution would be to make fedi a bit more like email: turn instances into access providers and have communities at a higher level that ignores instance boundaries. The important thing is to separate community concerns from provider concerns since having both intrinsically tied together is confusing and frustrating. It's also stopping us from having an actually good abstraction for communities because "we already have communities", except it fucking sucks.
My idea would be this:
There are communities which each have rules, admins to enforce the rules, and their own timeline.
Choosing your instance would work like with email: you compare some straightforward data points and that's it, but at the core it really doesn't matter which one you join. It's just an access point to the network.
Joining a community then can happen at any later time after you've joined an instance. The web UI lets you browse communities and easily switch between the community timelines.
Getting banned from a community would have no impact on your ability to communicate with people outside of it. You wouldn't even have to join a community, it's just a useful way to talk with people sharing your interests.
Notice how this is equal to joining multiple instances on our current fediverse to achieve the same thing (being a member of multiple communities), but now you only have one account and communities actually make more sense because they're focused on concepts instead of domain names.
Do you see the parallels with email and mailing lists which are essentially communities on top of email? It's exactly the same dynamic.
There's probably a lot more to consider, and I don't even know if it would be doable, but this is the basic idea I would pursue if I were to redesign the fediverse from scratch.