I guess it would be possible to make a 50% cancelling, but I don't think that's how they've done it, you wouldn't get that cutting feeling when peeping if they leave some AA-effect in.
It's important to note that a camera AA filter is quite weak too, it *is* possible to get moire with AA filtered cameras too, it's just a whole lot harder. So the manufacturer chooses some threshold where you get away with almost all moire and aliasing in almost all situations, but not not more than that so you keep almost all real resolution.
Then we get the interesting question - if the AA filter does not solve the problem 100%, how much worse is it then to be without AA filter? I cannot answer that question really so I'd like to test. How sharp lenses are and how effective the microlenses on the sensor are will have an effect. If there was such a thing as an aliasing-problem-scale I'd think that being with AA filter would be 95% problem free, and without would be 30%, ie there is quite a large difference, but say it would be 70 vs 60 percent than one maybe think the R would be worth it...
I do know that the microlens-less MFD sensors like 39 megapixels P45+ and indeed my H4D-50 combined with razor-sharp tech lenses will have more aliasing issues than the 5DsR, but without testing I don't know how large the difference is. Would I buy "in the blind" I'd surely go for the 5Ds though, for the principles behind it. 1-2% extra resolution is not worth the possible extra issues.
The other point is interesting, if those aliasing artifacts in fact is an important component of the elusive "MFD look". I don't know. I haven't really seen the "MFD look" and not really seen anything special about it. The older CCDs have both this pixel peep microcontrast thing but also quite high noise level so you get a certain "grain" or texture, while a clean CMOS with AA filter might look more "artificial" despite it's more exact in its capture. Possibly this is a "MFD look" thing, but again it's pixel peeping. Now when MFD has the superclean CMOS that aspect is going away.
I'm pretty allergic to pixelated prints, but that too is a matter of taste. I haven't yet had much experience of really large prints, but heard others experiment with adding some light high resolution noise on top of the upscaled image to get a pleasing texture up close on the finished print, and that's probably something I would look into too, but to make the upscale work well to start with the original picture must be suitably soft on the pixel level otherwise pixelation is very hard to get rid of. I don't like the look of fractal upscalers, creating more resolution than was captured is simply not a good idea. I've looked at a lot of upscaling algorithms and the basic "bicubic smoother" always gives the best looking result.
As noone really knows what the MFD look is it is tempting to mimic them as much as possible, and sure the 5DsR is more like them. But pixelpeeping an image to find the MFD look I think is cheating
The reason why MFD lacks AA filters is not fully clear to me, but if I may guess the reason is that the CCDs used, which are used for many other applications, did not come with AA filters in the standard product range so they would have to be developed specifically for photography probably at a steep cost, and then it was better to make it a feature. More resolution is easy to understand and see, while the advantages of a slightly softer image takes a bit longer time to explain... I think it became more obvious to me now when I've started with printmaking, you're always resampling the image to match the print size, that pixel peep sharpness becomes less useful and that aliasing becomes more of a pain.