Graphics programming for VR is only slightly more difficult than normal. Some screenspace techniques don't work as well, and everything is done in a stereo pipeline. The Oculus also requires some distortion and calibration of the distance between the eyes is very much preferred. Basically that's it.

Replicating the real world is still very far off, but there is an estimate where the resolution offered by the screen matches the average density of receptors in the eye. I don't recall exact numbers, but I seem to remember Michael Abrash having said something about it. Probably somewhere in the range of 8k displays per eye.