I have tried the "tape trick" before on assorted crop sensor bodies and lenses, but never had any AF performance with a f/8 system that I would say was usable. That's why my original comment suggested a difference between having any f/8 AF, and one that works well.
Of course, that's with an AF system programmed not to try to achieve AF narrower than f/5.6, rather than one which has been optimized to do so from a firmware standpoint.
Look carefully at Nikon's D4 AF specs. With f/8, 11 points including 1 cross; between f/5.6 and f/8, 15 points including 9 crosses. Does that mean the center AF point is an f/8 cross? If so, the shorter baseline means Nikon traded away accuracy. What about the 8 other crosses that don't work at f/8, but do at, say, f/7.1? Are those f/7.1-sensitive sensors? Without seeing a schematic of the sensor or more details from Nikon, we won't know for sure, but I suspect the answers are no and no. I think they'd not choose to sacrifice accuracy, but maybe the sensor is just more complex, and where Canon chose to squeeze more AF points and make more of them crosses, Nikon chose to squeeze in some f/8 and f/7.1 lines.
Still, I think it's possible that Nikon is using the same f/5.6 AF points, and just optimized the firmware to support narrower apertures. If so, it's possible for Canon to do the same (although there may be non-technical reasons that they won't).
I'm not so sure ... there's comments somewhere that Canon changed the sub-mirror responsible for autofocus from being elliptical to flat. If that's actually the case, then it may be that there's less "concentration" of light with the new AF mechanism and thus less ability for AF to work with less light.
I don't think that would matter at all. The aperture limitation isn't really about the amount of light, but rather about the baseline of the sensors. An f/5.6 lens won't activate f/2.8 sensors because the sensor lines are widely spaced, and the narrower aperture doesn't allow the light to be spread far enough. Consider - if it were the amount of light, AF would fail with a 3-stop ND filter on an f/5.6 lens, but instead it works fine, as long as the intensity exceeds the EV sensitivity of the sensor (-2 for the 1D X, -1 for previous 1-series, and -0.5 for other bodies).
Also, 'concentrating' the light implies the mirror was concave - I don't think that's what is meant by oblong. Rather, it was more oval rather than rectangular. Going from oblong to rectangular for the 1D X makes perfect sense - look at the shape of the AF point arrays - oblong on the 1DIV and 1DsIII, rectangular on the 1D X.