I relay don’t know how relays will handle real traffic. some of them are already overloaded
Discussion
There’s likely a ton of optimization to be had at the relays. Things like caching, read replicas, and so on.
It's eventually going to take relay software that can scale horizontally. Not just be optimized vertically.
I’ve been thinking a bit about this. I think the protocol and relay implementations lend themselves to horizontal scalability. Just a hunch but the db models I’ve seen are very simple so it’s likely easy to setup some very good replication with the DB tools we have at our fingers. I do kind of want to try it just for kicks.