There is no way you could depend on the latency being regular enough to rely on it, its going to fluctuate over time.
But if you forced the delay to a manageable frame.
I can see a path and without using quantum theories, but its complex.
Each input (singer, guitarist, drums whatever) plays and transmits an embedded timecode with the signal. Someone with the right skills builds a system that captures everybody’s audio along with their timecodes, and syncs them all up and allow say 3 seconds to account for most circumstances, then sends the mix to each player minus that players part. And sends the entire synced mixed result to spacial audio. There would have to be a separate audio channels to send to each player and returns for each player.
That player hears everyone else in sync and has to mix his own sound with the delivered sound locally, to jam with, and doesnt listen to the server sound, and only sends his sound with timecode to the synchronizer.
Like a compressed chain, so to the audience you are live, but to each player you are playing to what seems like a pre-recorded track and there is no immediate human feedback. If you want to change the groove its going to take a few cycles.
But there’s the software development to think about, and the channels. There’s a project for someone, shall I put it on the worklist? lol
I’d like Judas’s idea about waiting 12 bars then coming in, if it wasnt for my hatred of 12 bar blues lol.
Could pick any song that just repeats a short loop and just come in where you feel like it, I wonder how that would sound to the listeners, I would prob still be out of sync because of my own latency.