Very thoughtful points. One thing about nostr is that it does tend to balkanize due to the technical architecture, allowing for different groups of people to use it in different ways (different relay policies, client features, filtering, etc). But the tradeoffs you list are real, and enforce real constraints (the biggest of which is bare keys as identifiers). Many of these constraints can be designed away, which keeps me optimistic. We've had 30 years of research and development into password management, but far less into end-user key management. Even if nostr itself has some fatal flaw, I think a lot of interesting ideas are coming out of it, just because it provides a very different set of affordances for digital spaces.