afni.nimh.nih.gov

I'd give that a definite maybe. :) It makes sense, but
automation can be dangerous.

This only matters when the outcount value is high but the
motion change is small, say if motion correction could not
improve the image correspondence, but the correspondence
is still bad.

It certainly seems justifiable, but I have not spent enough
time playing with such data to advise one way or the other.

Sorry for the wishy-washy answer... :)

- rick