What I mean is that typically users send out little bits of data (which are each individually signed) and then the clients aggregate them into different views (for example, the list of people I follow, or the list of people following me or the list of replies to a post). In most cases, the user just sends out the granular action (zap, heart, reply) and it's up to the clients to query the relay and build up the aggregate list. But with the list of follows, it works different, the user keeps republishing the FULL list and the relay treats it as replaceable, it just seems inconsistent with how nostr typically works.

I think "Sending little bits of data and hoping they will somehow be reassembled correctly on the other side" is actually EXACTLY how Nostr works and that's a good thing!

Yeah I was thinking something like https://github.com/vitorpamplona/nips/blob/relationship-status/81.md I'm definitely new and still learning the protocol in's and outs!

I get that, seems like the whole system is moving fast and evolving organically which to me is promising. I guess there is something like https://github.com/vitorpamplona/nips/blob/relationship-status/81.md which seems like it could supersede a simple flat follow list. How do these things kind of move into widespread use? Is it a matter of enough popular clients adopting the new protocol?

I think just publishing the individual follow / un-follow (delete follow) is consistent with how most things work in Nostr. If I want to see all the replies to a post, I have to query relays to assemble that list, same to see followers (as you mentioned) or reactions to a post. It is a mess in some sense but I think it's a good model.