Although it seems like the only reason for four on the front is so that you can block two with your hand and still have two left for distance finding.

Correct. The four front-facing cameras are low res, b&w, low framerate.

ETA: I worked on this feature for lab126.

Any chance apps can use raw imagery from those cameras? can imagine some pretty nifty uses (say, true live 3d videophone: use cameras to construct 3d model from low res data, then map color image onto model, just have to transmit already-sent video feed plus modest model data).

The API is open as of today, and I suspect you only get information on the events (looked [direction]; peeked [direction]), not raw data.

Interesting! B&W makes total sense. Low res also makes sense for head tracking but I guess that means pupil tracking is out of the question. Regarding low framerate, like lower than 30 fps? I would have thought continuous perspective shifting would require a framerate at or better than screen refresh rate to appear normal to the user.

You can't see pupils really. Watch the videos they showed behind Bezos and they show the test apps with raw video.