I know you’re playing devil’s advocate, but to play devil:

In a theoretical world where you can manage to perfectly beamform the entire 20-20k Hz frequency range into a single node (or pair of nodes around the ears)… you’re still just re-condensing the original reference reference signal at the site of your beam target.

And if your idea of peak quality is to hear the reference signal loud and clear, it might be marginally easier to set up some well-tuned speakers in an arrangement relatively free of resonance hotspots and then crank up the volume.