In practice, this can't actually happen because what if you need to play a notification sound? Then you would need two different streams to send to the device without having to reencode everything.
I can't think of any technical reason why companies cannot work on changing their audio stack and adding to the next Bluetooth specification to support this scenario. I am not sure what would be the benefit to doing this work through, other than saving encoding stop on the device side.
This but in general this isn't how audio APIs are structured. Some apps would do their own decompression, for example. Some may want to do some additional processing on the PCM data as they play it. In general, all related APIs and apps are built with the expectation that decompression happens locally on the device. It would be a massive undertaking to change that.
If I remember correctly Bluetooth Dual Steam is a thing on latest Profile.