I am not saying whether sending happens at the establishment phase or not - it's just that when your AV capturing is started and even encoding goes on - it's a matter of dropping the actual AV packets or sending them at the network stack/jitter buffer level. By the way, if not for privacy/security, it's actually very useful to start sending AV stream over the network, to pre-heat the network & test its throughput. For example, cellular data bitrate ramps up only when you actually send data, and it takes some delay to ramp up at more power-hungry levels and for cellular to even test what's possible. Also, estimating network bandwidth for the application layer requires measuring the round-trip time for a few seconds...