I don't know how many participants, I was thinking about small groups (<10), a typical friends and family scenario. Maybe the color code could be activated if the user's activity goes over a threshold, otherwise it stays gray; so there are fewer risks of overlap colors and messages from usually silent users are easily spotted.
No, I didn't think about threads. On mobile, probably a linear structure, just quoting the message I was replying to, is more usable.
I don't know the Cezanne theory, I will study it, thanks for pointing it out. Btw, it makes sense, warm/vibrant color naturally signal activity, while cold/desaturated ones convey the opposite feeling.
Nostr gives us the unique opportunity to test different UI on the same data, the perfect lab.