waxwing on Nostr: I'd say it's a better scenario than "prove your encryption algorithm is secure", ...

I'd say it's a better scenario than "prove your encryption algorithm is secure", although still hard (to prove). The former is essentially impossible but as you say "test of time" at least moves you probabilistically towards proof. The latter "prove that the system is censorship resistant", I would say, is better, even if not possible to define fully formally, and both of those things are because "the censor" is not a clear concept. There are certain powers that may have the ability to censor with the threat of violence, in different geographies, different times, different corporate entities etc. And they are capricious: they might ignore nostr for 4 years and then suddenly be 100% focused on it. So it's only better than the encryption case in that, when the attack really comes, it will give very clear evidence whether there is an effective censorship resistance or not, but that attack may not come for a long time (or never, also possible if nostr is not hugely successful).