Thom Hogan responded to me once that IBIS left the AF sensor with an unstabilized 'view'. With its 'view' jittering and twitching all over the place, how can the job of AF and tracking be anything but extremely difficult?
I get the appeal of this idea but I'm not sure it stands up to scrutiny. It's fair to assume that PD AF samples the scene at some rate of "looks" per second, and that some information derived from each look carries forward through time. Does the "jittery IBIS" hypothesis assume that the information that carries forward is about the location of relevant detail along each sensor array? My guess is that each look has to be processed independently to derive the focus error between the detail seen by pairs of arrays (since detail from a moving subject can move across an array from look to look regardless of whether the view is stabilised). The relevant information that needs to be carried forward is the focus error, which is about the distance between detail over two sensor arrays rather than on the location of detail along the arrays.
The one thing that might be credible about the jittery IBIS idea is if a look takes long enough that the "image" seen by an individual array suffers from motion blur, thus smearing the detail. In the absence of evidence that that's the case, that it's significant, and that stabilising the image makes a significant difference, I wouldn't make any conclusions about it.
