As for ChipMixer, I believe they have very good reasons not to implement Bech32 yet:
I'm not sure I follow that logic. The anonymity from using ChipMixer does not rely on an attacker not being able to tell an output came from ChipMixer. Indeed, it is trivial to identity a ChipMixer output, given their characteristic funding transactions with 50 outputs of 0.016 BTC, or similar. It is therefore irrelevant if an output is legacy or Bech32 - it can easily be identified as a ChipMixer output either way. Rather, the anonymity comes from being unable to link these outputs to any inputs due to the time travel funding structure and set chip size.
Thanks for pointing that out. I may have been approaching this question from the wrong angle.
Two questions for you:
If ChipMixer were to, in practice, exclude users who haven't upgraded to Bech32-compatible wallets, how would that affect the anonymity set of the overall service? Would it matter, in your opinion?
When you say an output could easily be identified as a CM output, I assume you are talking strictly in terms of an heuristic analysis, and not necessarily in terms of blockchain analysis, correct? That is to say, if I merged and split my own outputs exactly as I have observed ChipMixer doing and then sent you some of those chips, you would assume they came from ChipMixer.