This tries to merge multiple mics into one output on multiple channels.
https://sirdarckcat.github.io/binaural-recording
Proof of concept works, but requires calibration. For calibration it could use gcc-phat (see /eear) to calculate the shift from the mics https://github.com/sirdarckcat/sirdarckcat.github.io/commit/effa8dd35fb72ae8c1abad11c4531feeb7bb3da1
Then it may be possible to apply a delaynode to reduce or eliminate the delay on post-processing of the wav.