Parsing on input keeps the network sane and helps everybody. You do it or not do it, it's your client's decision, self-contained and virtuous.
Parsing on on display is Postel's law and trying to fix people's mess while incentivizing them to do more and more mess until you can't fix it all. If you do it you create work for everybody else and protocol bloat.
I think 1 is completely unnecessary, and to go against this just a little nostr:nevent1qqsqqqzz0ckjlmwpxcp8999rahl06pz5tq4xywnrcdk7jjkx5p2uuxcpz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsygpm7rrrljungc6q0tuh5hj7ue863q73qlheu4vywtzwhx42a7j9n5psgqqqqqqs8fpf3k which I very much agree with, anyone can include any markup in any note to be rendered how the client prefers. We already do this with URLs. I can see no good reason to encode this behavior into the protocol.
As far as 2, I think the architecture of the network does ensure censorship resistance, but youre right, it cannot be consistently verified for all notes, and so the claim isn't as strong. I don't know that a chain of events is a good solution, or even that a solution is needed, I'm pretty convinced again by the claim in reference to client relay architecture and signed messages, and not entirely convinced that the added complexity is worth making the claim stronger.
