The way that I see it, its not just a technical problem anymore. It's about making the methods as diverse as possible and to some extent messing up the network for everyone. In other words, we should increase the cost and the collateral damage of widespread censorship. As an anecdotal data point, the network was quite tightly controlled / monitored around 2023 in Iran and nothing worked reliably. Eventually people (ab)used the network (for example the tls fragments method) to the extent that most of the useful and unrelated websites (e.g., anything behind cloudflare, most of the Hetzner IPv4 addresses, and more) stopped working or were blocked. This was an unacceptably high collateral damage for the censors (?), so they "eased" some of the restrictions. Vless and Trojan were the same at that time and didn't work or were blocked very quickly, but they started working ~reliably again until very recently.

[1] https://www.petsymposium.org/foci/

Here's an overview. Be warned, the conclusion is:

> We enumerate the requirements that a censorship-resistant system must satisfy to successfully mimic another protocol and conclude that “unobservability by imitation” is a fundamentally flawed approach.