Regarding convergent evolution in the great apes, I thought it was well demonstrated that knuckle walking was convergent, because the mechanisms for spinal stabilization are distinctly different between gorillas and chimpanzees - and orangutans, who also use their palms instead of their knuckles.
See for example the following article; figure 26 illustrates how orangutans and gorillas stabilize the spine through locking of different parts of the spinal vertebrae, while figure 28 shows how Pan achieves its stabilization through a system of ilio-lumbar ligaments.
Given convergent evolution of similar locomotive behavior, the wrist features almost have to be convergent, and convergent evolution of morphological features in the hips and spine shouldn't be surprising.
You're correct that there's a good argument that the chimpanzee and gorilla forms are non-homologous. I am inclined toward that point of view, also.
However, a lot of people are unpersuaded by those observations. Chimpanzees and gorillas are very different in size, and it would be surprising indeed for them to carry their weight identically in every detail, as their functional requirements are different. So we shouldn't expect them to be identical even if they retain knuckle-walking from a knuckle-walking ancestor. Williams (2010, doi:10.1016/j.jhevol.2010.03.005) argued that independent evolution of the hand and wrist traits supporting knuckle-walking is unlikely given the lack of morphological integration shown by the variation within chimpanzee and gorilla populations. That argument doesn't go too far with me, but it does suggest that the similarities are not an easy parallelism but a hard one for selection to generate.
The orangutan and gibbon convergences carry a lot of weight with me, as it seems clear that the common ancestors of orangutans and the rest of us were quadrupeds. As you mention, that's not a knuckle-walking issue, but goes to limb proportions and lumbar spine function.