While we're at it, it'd be great if we could avoid remuxing e.g. facetime audio, which is AAC, and notification sounds, into a single stream before sending it to bluetooth. Would be nice to avoid the latency and just shove the raw AAC from facetime into the headset, and when a notification ping arrives, send that as a separate audio stream with maybe a different codec
Yeah, the ultimate Bluetooth audio protocol would probably be a meta-protocol, combining the two ideas I mentioned with a MIDI-like timecoded sequencing protocol. You'd pre-buffer one or more notification sound effects onto the receiver, registering them with audio-session-specific IDs; begin streaming the live audio (the stream gets an ID); and then use the control sequencing stream to say "mix in a copy of registered-stream N [the ping sfx] at time T." (And the sender then cut the sfx off early with another such command, if it wanted.)