I agree that metadata privacy and censorship resistance is hard, because it usually means you end up relying on something like a DHT system where everyone can see the origin and endpoint of a message. Arguably putting envelopes inside of envelopes, metaphorically speaking, can obfuscate this a bit (and this is essentially what you get with the Tor network to a degree, as each node just knows to address the next node, and that it came from the previous one), but usually, you end up with leaky metadata, as encrypting this addressing information en route is hard to do secure key exchange with.

As for privacy of contents, however, that shouldn't be hard to do while also being censorship resistant. PGP was around in the mid-90s and the concept still holds. 

Seems like "privacy" is too broad a concept, as it can apply to the confidentiality of contents, the anonymity of the sender, or the anonymity of the receiver. And at some point, if you can somehow get all of these right, you get another issue: authentification, such that the receiver can identify who it was sent from, without anyone else being able to do so.

My head's already hurting and I'm not an engineer, but I am inclined to think that SimpleX Chat does have some 'splainin' to do because their assertion doesn't seem to obviously follow from any agreed upon premises at this point.