This sounds like a good idea.

I keep thinking that at Twitter scale, total data will be north of 20TB per day and nobody will be able to store all of that, but if let's say 50% of all users run relays and each of these relay stores the user's very own events (obviously) but also all events of let's say up to 100 or 1000 users that this user follows, then the amount of data that needs to be stored per relay is manageable and everyone will have a huge amount of backups.

But then, I assume, it will become impossibly hard for each client to properly fetch all events that a user should see on their timeline? A client would have to send out hundreds/thousands of requests to get all the data that is scattered accross tens of thousands of relays??

I should probably sit down one weekend and read the protocol spec and try to implement a client and a relay 🙈

nostr:note1hzwn3829ctjh6xyc4kmedynvyf4zaxldhlj733fe9zp9z5e0q4vqxjedf2