IMHO blastr was the right solution for nostr early on. It was immediate (the outbox model was longterm), it fixed a connectivity problem that nearly all clients had, and it didn't need clients to write any code (the outbox model requires client support). Nostr would have been utterly hobbled without it, unable to grow beyond a handful of relays, and I'm sure that adoption of nostr would have been severely impacted as well.

Any suggestion I've ever made about blastr (or other forms of event copying) being "bad" should be interpreted as "we should try to move past the need for event copying," not that event copying was bad while it was needed.

I've always believed event copying was the wrong long-term solution. But it was absolutely the right short-term solution.

Also, I think clients that have their users configure a fixed set of relays are useful clients for some purposes. For global feeds. For chat. For community relay situations. Even if they work in concert with a client proxy.

It is just that they aren't good at following and replying to anybody that the user wants to follow and reply to... only that subset of users that use the relays which are configured, or if events happen to be copied into those relays. And this is what fiatjaf, pablo, hodlbod, even jack were trying to make obvious. Put users onto small relays where events aren't being copied from, and suddenly the shortcomings become more obvious.