It actually turns out that it is possible to have that working. It was enough to start the app with longFormAudio, and then immediately switch to mixWithOthers. This makes it possible to play in the background and mix with other apps (including spotify). I still think this is somewhat a backdoor approach and not sure if it's the intended behavior, so it would be very helpful if this was properly supported or documented.

Thanks for the reply! I have also tested AVAudio​Session with .playback + .mix​With​Others on watchOS. That works for mixing my metronome with Spotify while the metronome app is in the foreground, which matches the behavior I already see on iPhone. However, on watchOS the metronome stops once the app goes to the background, while long​Form​Audio keeps it alive but then conflicts with Spotify. So this seems to be a watchOS-specific limitation rather than a general AVAudioSession limitation on Apple platforms. It would be very useful if watchOS supported a way to combine background audio persistence (long​Form​Audio) with cooperative mixing behavior (mix​With​Others) for lightweight apps like metronomes, cues, or coaching audio. Is there any way I can formally request this feature other than the FB22273077 I have already submitted?

Also submitted a feedback here FB22273077