Indeed some ground work has already been done when we introduced the NDI nodes. To my knowledge the current high level MediaFoundation API we use to grab the video frames doesn’t give us access to the audio data. I could be wrong here, would need to check.
However there would also be a different approach with a lower level API provided by MediaFoundation where one has full control over the playback by manually grabbing the samples (video frame / audio buffers). Some tests were done already with it just with video (video player, source reader). It was simlpy playing back videos as fast as it could. Changing that sample to fetch audio samples should be doable. The hard part however would be to get the synchronization right. Would be fun to get into that topic once more, but as always other parts in our universe need attention as well ;)