Makes sense. Totally agree. From my point of view we have three options:

1. Make nostr more about the self-validating event structure rather than transportation method. Normalize sending nostr events over different channels, like Tor relays or bluetooth like Samiz ( https://github.com/KoalaSat/samiz )

2. Find more incentives to run more relays, scaling out rather than up, more smaller relays. Harder to censor entirely if we have hundreds of thousands of relays and clients are good at crawling to find the right ones ( i.e. outbox model https://www.whynostr.org/post/8yjqxm4sky-tauwjoflxs/ )

3. Instead of focusing on censorship resistance, Nostr focuses on something it is already very good at - data interoperability. Connecting lots of data to each other, making it useful across a number of apps, unsilo-ing the web.

While number 3 doesn't really address censorship resistance, I can see a scenario where Nostr uses this peace-time (where nobody is being censored) to become so valuable censoring it would cost more than it is worth. Like if a government today tried to shut down or censor the Internet, it would have massive economic costs to a large chunk of the economy, being so unpopular it could potentially stir an uprising.