I spent some time this week writing a new key rotation scheme for nostr:
https://github.com/nostr-protocol/nips/pull/2114
I think it's pretty good for what it is (it improves on nostr:nprofile1qyxhwumn8ghj7e3h0ghxjme0qy2hwumn8ghj7un9d3shjtnyv9kh2uewd9hj7qg3waehxw309ucngvpwvcmh5tnfduhsz9mhwden5te0wfjkccte9ec8y6tdv9kzumn9wshszxnhwden5te0wpuhyctdd9jzuenfv96x5ctx9e3k7mf0qqs04xzt6ldm9qhs0ctw0t58kf4z57umjzmjg6jywu0seadwtqqc75s49cdh4's version by not requiring an arbitrary time delay). At the same time, it's riddled with very fragile assumptions about events being available, introduces a hard dependency on OTS (or some equivalent), and requires clients to constantly compute key validity.
At the end of the day, what I have learned is that we probably can't realistically do key rotation on nostr in the application layer — we need some kind of cryptographic magic or strongly consistent data store to make it happen. It still might be an interesting read for the nerds out there though.
