> I find it hard to believe no one has requested Content Aware Fill for AE in the years Photoshop CS5 introduced it.
That is the wrong conclusion to jump to.
Algorithms that are easy (or even trivial) for still images are often much, much harder for sequences of images (i.e., movies), since still-image programs can make arbitrary decisions about image processing without considering how the same decision must be made for a very similar frame that is adjacent in time. Just think of all of the bad rotoscoping work that you've seen where edges "sizzle" because of small deviations from frame to frame... even though each frame seems fine on its own.
Content-aware processing for video is a very rich area for research, and our best people are on it.