I'd like to start with some comments on discussions I've followed in the forum and I do it here, so it is comprehensive to read for ppl that are not reading the forum otherwise and easy to find again, additionally.
One suggestion or thought which is coming up repeatedly is about engine switch. For CQB I can clearly say no. There are multiple reasons and you can also be ensured that there was much consideration from my side before I came back to W:ET again. One main reason to stay idtech3 based is that the existing code base makes it possible to get something like CQB done at all. Never forget, there is like 10 years experience and countless man-hours involved and I know every line of code by heart, so to say. Another issue with modern engines is the huge workload required to get decent bump/specular textures and many other assets done that would be on par with the AAA quality of plain color-channel texs we have now. The third main reason is the existing very well known production pipeline including (most importantly) the map editor and the md3 model pipeline.
That brings me directly to the question of how much is doable with that old battle horse idtech3? A lot. You have to feel comfortable with the retro look due to lack of bump/specular, though. - As a side note, I have to admit I started seeing it as a feature and it gives a good feeling to be so far off the current games that the hunt to somehow try to catch up is finally over. - So, what's doable? CQB is aiming at 50.000 - 60.000 verts per scene @ 60 Hz, which on average compares to a similar amount of triangles. At my machine (3 GHz Core2 and GTX 275) this runs with stable 100-110 fps. It still runs well on my aged Laptop with a NV 7400 Go (can't remember the fps right now). In the years to come, most ppl will be able to have stable 60 fps with very rich graphics this way. So, were would the limit of idtech3 be? There is no hard limit for polycount, the problem however is the architecture of idtech3 which misses vertex buffer objects (VBO). As such, there is always the bottle neck of compiling vertex arrays and sending it to the graphics card on a frame basis. As hardware will be faster and faster, we probably can do 200.000 polys at stable 60 fps in just a few years. The issue is more that it could be done today already if VBOs were available (see Xreal and other idtech3 beef ups).
Much talking, the bottom line is that excellent models and excellent textures with decent poly detail in stable stock W:ET is much more than buggy beef-up engines with old maps and strange wanna-be bump/specular maps.
Finally, a friend uploaded a camera screen capture showing off CQBs new free look feature. It's not best quality but you can see some action and how CQB will look. Thanks for that and enjoy: