Right back at ya! I’m not trying in insult ya. We have a problem here tho. So, back on topic.

Ok, few issue you may not have considered are:

Non persistent note flag is implementation specific; if my modded relay is set to snoop all your traffic and record traffic - Intractable problem. Open source and voluntary nature of nostr relays. Not an issue with tor, due to telescoping (eg: intermediate nodes on path do not see routing metadata). Even if we encrypt the full payload, the npub is in the clear. This is not the case for Tor.

Routing metadata - tor vs relay; I appreciate that there similarities with peer routers and guard/tor nodes - however the telescoping behavior of Tor circuits leaks much less metadata. I’m already assuming I’m capturing all the notes, and now I’m snooping on the addresses too. The behavior of the intermediate nodes on the path is a known quantity too vs implementation specific behavior in nostr relays.

Other have alluded to latency, which apparently I know nothing about. There is a publish, relay, database cache, and fetch cycle associated with each note which does not exists over a TCP/IP or Tor circuit. It’s not to say you can’t design a low latency solution using the protocol, but do not ignore this trade off. I would estimate the order of latency for note propagation vs single LN message over Tor to be ~10x. Happy to be proven wrong here. It’s comparing a broadcast application layer protocol (in JSON no less) [nostr] to a low level network routing protocol [tcp/ip or tor circuit]. Of course it’s going to add trade offs.