"couldn't 4k x 2k improve that some?"
As Abrash said, 8k x 8k is about ideal. You're looking at a 128 megapixel render instead of today's high-end 6MP. At absolutely optimal Moore's law, that's 7.5 years until we gain that capability. Given that we've been making gains of significantly less than 2x / 18 months in performance in GPUs, focusing more on cost and efficiency, we are likely 15 years out from achievable 128MP rendering in real time. On top of that CPUs have stalled. We've roughly doubled performance since 2004. Far less than the roughly 16x performance improvement from the 8 years from 1996-2004 and 1988-1996.
There is also another enormous problem. Small, high-density 720p panels for the prototype, and tiny 1080p panels for the shipping unit are available because of cellular phones. In the two years since the iPhone 4 hit, there has been a huge push for high-density displays, making them plentiful and cheap. The problem is, Apple called their display "Retina" for a reason. Now that we have almost universally enabled cell phones to exceed our eye's ability to resolve pixels, what is the push to move resolutions higher? 960x640 at 3.5" or 1920x1080 at 5" is as sharp as could be needed. Any extra resolution is throwing processing power, and therefore battery life and weight, at a solution with no benefit. 4k x 2k doesn't look any sharper than 1080p at 5" when held a foot and a half from your head. Neither does 16k x 8x.
So, without cell phones pushing resolution any more, that leaves VR to be the sole pusher of higher density displays. That quickly creates a "chicken and egg" scenario. Nobody makes VR displays because there are no suitable panels. Nobody makes suitable panels because there are no VR displays.
Let's look at my living room now. 10 years ago I had a 27" CRT. Today there is a 64" plasma sitting in there, for not that much more money. The days when we can get flat panels that cover an entire wall are rapidly approaching. I only have four walls to cover. I also now have a motion sensing eye in the room (Kinect) that people have already built solutions for involving full body tracking and display projection within a space. In short, full-room display and tracking is a lot more realistic of a goal, and a lot more viable as to how we get there without any sudden leaps in corporate altruism.
Carmack, Abrash, and Palmer are all brilliant. They are far smarter than any of us here. I fear that they all may be having difficulty keeping realistic expectations due to project proximity.