GY,
This seems overly complicated and unnecessary given the low cost of solid-state memory. In fact the idea that the 'recorder' is manipulating (mixing and storing) the data once recorded seems astonishing to me... but then I suppose the design is some 20 years old or so (haven't checked that - it is a guess!).
It seems they designed this "architeture" to allow some flexibility in the a/c config. (microphones, etc.).
But i agree with you. IMO this shows there are "room for improvement". For example, why not a better fidelity in ALL channels (Nyquist) to allow an easier analysis. Memory chips are cheap like you mentioned.
On Pingers we will "soon" see in the market solutions that could avoid this ABSURD two years (3rd Interim report by summer) delay.