IMO this probably isn't just about latency. keeping people in voice gives them training data text never will. is that why they were fine going transceiver over

IMO this probably isn't just about latency. keeping people in voice gives them training data text never will. is that why they were fine going transceiver over sfu and mostly ignoring multi-party?