Supposedly this was something the Nvidia devs were not necessarily expecting to see but happened to stumble upon it. In other words, it arose naturally from their treating light as physical rays rather than being programmed in, and if so I think that's pretty neat.
Is it fair to say that the image is always there, but lost in a wash of additional light, and that a camera obscura filters out the other light, revealing the image that is already there? Or does a camera obscura somehow shape the light in a new way to produce the image?
http://fslashd.com/2015/04/circles-of-confusion/
http://www.villageinframe.com/how-big-should-the-hole-be-in-...
This is a useful image. https://qph.fs.quoracdn.net/main-qimg-e92438d691243116d52f99...