hmm now that I think about it it wouldn’t be hard to incrementally move to the gossip model in damus. I could calculate a subset of the fastest relays to query based on a thread event author and the relays they have configured. I would still be concerned about missing events but maybe it would be fine.

Might be better for people with huge relay lists? I still feel like this is a premature optimization though. Relay pools are dumb but simple. I would have to see if the gains are worth it for the complexity.