Many small holes collapsing into a few large ones, perhaps? You can sometimes see where two holes merged.

That's a good idea, but if the holes collapse I expect more variation in size. My guess is that the CO2 diffuses until it finds a nearby hole.

Did someone put a whole cheese in MNR to track the holes? (I guess an ultrasound image device is cheaper. Is it possible to use a CT adding contrast to the cheese?)