Can someone provide/link-to a principle of operation? I looked at references (https://www.dragonflytelescope.org/publications.html) but they point to publications that look to be behind pay walls.

In particular, I don't see how N co-aligned cameras is any different than N images taken in sequence with one camera (averaging over noise), other than a reduction in time required to take N images with one camera.