I hope this is accurate as the old 18-35mm was easily the worst lens I have ever owned. I tried two, both from different batches and each managed to grow light blobs on highlights! Parked cars looked as though their headlights were on and resolution, even on a D200 was sepulchral. Now you are claiming the new G lens betters under all use the old 17-35mm f2.8, but the old 17-35mm f2.8 was said to better all the primes at the focal lengths it covered. Here I note the resolution you get from it, as with the 17-50mm AFS f2.8 Nikkor is dreadful - practically everywhere and all the time, especially as you go further into the longer focal lengths. Am I crazy, but maybe just possibly this is due to focus errors?? Are you using live view. Are you, IN FACT testing all these different machines as you claim, with every single lens, or merely extrapolatuing sharpness and other scores from a basic actual reading- as gas and electricity meter readers estimate performance??
What makes me suspicious is that my 17-35mm lens is rather better than you tell me it is on a D800. Interestingly I am getting in an old 24-50mm f3.3-4.5 lens to use. Notably it is ignored these days, but it was, for 18 years Nikon's ONLY 24xxx zoom. Surely despite reports, it must be OK? as that is a very long time for the best-known manufacturer to produce only one model zoom.
I have also bought and returned 2 24-70mm lenses as they were only ever sharp in the middle due to field curvature. Even at f5.6-8 on a D800 they were a joke- and your test concurs only as far as chromatic aberration is concerned. Your tests do not reflect actual use- which is a shame as I am sure a practically bad lens might be bad at usual distances, yet fine in front of a test target.