depending on the density of coverage with the other inks.
This makes me realise there is something I don't know, presumably of academic interest, but who knows: to what extent does total ink density vary?
To take a simple case, a BW image with a single black ink would vary ink density almost linearly with optical density. However once you are mixing 3 different shades of grey/black, the situation will be very different: I would guess the total ink density would be reasonably constant down to the tone corresponding to the lightest of the greys, then drop linearly below that. Presumably the situation is similar but more complex when mixing CcMmY, plus a couple of primaries, plus the greys...
Which leads to question B: does the CO taper down with total ink density, by trying to cover only the actual points of ink; or does it attempt to "fill the gaps", in which case CO would be anti-correlated with ink density? Since we know CO Auto does not coat paper-white areas, it seems more likely to me that it correlates with density, ie the first option. If that is the case, I wouldn't expect a dramatic change in the ratio of CO to total other ink. If otoh it anti-correlates, then the ratio would swing dramatically, with low CO use on saturated images (like Keith's pano) and high use where the image is low in saturation but not paper-white.
Hmmm.
Has anyone printed a fine mesh of white on black, to see if auto-CO switches off even for very small pieces of 0%, or does it only switch off on larger burnt out highlights?