I knew someone working in VR about 20 years ago, and the big problem then was latency - if there's too much delay between when your head moves and the scene is rendered at the new orientation, it just doesn't work. IIRC the eyes are sensitive, in this respect, to delays of less than a movie frame's 1/24 sec.

Carmack is both very concerned and very knowledgeable about latency in rendering, so it seems likely that he's solved this problem... but I haven't seen it explicitly addressed in anything I've read about his recent work. It might help his advocacy if he explained why this tremendously exciting thing failed last time, and that he's solved it, to overcome the "VR winter".