the ‘mother host’ (which is the menus, so what you refer to as OS) currently doesn’t touch the audio, and would need a bit of an overhaul to do so and also take more cpu.
(processing audio requires process to run at higher priorities)
in its current form, i would not undertake/recommend this…
where this might become possible, and more interesting, is if the ‘mother host’ started using libPD, this would mean in would run the PD patches directly, without running the PD process. this would be an interesting development, as it could improve efficiency, and also integration… and then because it would be processing audio, adding features like audio monitoring would make sense.
for the more immediate future, there are other ways/techniques to help patches be more integrated and consistent, which I’m currently exploring