once you use the audio expander, you basically get either 6 (SR 44.1khz) or 10 (2 44.1 8x 22kHz) audio IO. the additional ins/outs appear just like the main ones, so in Pure Data (for example), they are adc~ (1-6)/ dac~ (1-6) ... just like you get from a soundcard on mac/pc.
so yes, from here its trival to write a patch in pd/sc/c++ that would mix the 6 inputs, and send to the 6 outputs. (just like Axoloti, you can multiply the audio signal to reduce it, then add together)
And yes this could be controlled from midi
actually, ive been asking questions around this on the bela.io forum recently, since I had a similar idea... basically having 6 inputs, being fed into some dsp processes for fx, then routing those in mixed forms to the 6 output... so I kind of matrix mixer 'with dsp'
anyway, you'll see those questions have highlighted some interesting nuances coming up
there we have been discussing the SR criteria, and what drives this... its basically the PRU code. which the devs are now working on to be a bit more flexible. the next version will allow you to have 10 44.1khz inputs, but with no additional outputs (i.e. you just have the 2 mains outputs)
note: the audio expander uses the analog io for the audio io, so you 'lose' the analog ins/outs, again see my recent posts for some details on this