It's definitely getting closer to that point - for example winnie the pooh is a popular meme used to make fun of uncle Xi but it's hard to police given that it's a children's cartoon. It hasn't stopped the aggressive censoring of pictures of winnie the pooh (no joke) but it has definitely presented a challenge. Using vague historical anecdotes and rarely used imagery is a traditional Chinese mode of communication - in modern days you can see it in action via the aggressively close reading people do of CCP mouthpiece newspapers - the use of certain terms signal a greater ideological shift which people will then try to adapt to/guess at. Personally I'm curious about how this mode of communication will evolve given the current development of automated policing algorithms which will undoubtedly take context into account.

The Chinese already do that with regards to ‘winnie the pooh’ in reference to their leader (and when that was blocked, further obscure poetic language).

I think the govt. is happy with this change, because it dramatically slows down communication rates if people have to have built-in encryption/decryption parsing in their speech.

Also, if you accidentally inform a government agent of the code, so to say, you disappear and the in-group has to change their coding language again.

I'm wondering if we could develop some kind of encryption / steganographic system that would hide the true meaning of a sentence behind somewhat sensical human-readable text to lower suspicion and automated detection.