Just to state the fairly obvious 
If you print at the same maximum size as many do on e.g. an Epson 3880 (max super A2) then there is no difference in how critical focus is or DOF. If you look at pixels in 1:1 in Lightroom then you and judge on a screen like the retina screen on the MacBook Pro 15" (220 PPI) and print the final print will be about a meter or so in width (with no cropping) from the 5Ds R. If you judge DOF from 1:1 on a screen with much less density like a large 30" screen at the old default of 2560x1600 then 1:1 would correspond to printing so large that you probably never would do that. Then on top of that you would need to judge what the CoC would need to be for a normal viewing distance. For a 100 cm wide print would you view this as close as you would view the screen? Probably not so the requirement goes down.
As previously mentioned I shoot with my 5Ds R focussing using the PDAF in the camera and for landscapes absolutely no issue with accuracy. Compared to the 5D III I don't see this camera to be more difficult to shoot with despite the difference in resolution from 22 to 50 MP.
The attached example which was a casual shot from yesterdays morning when out searching for the light was shot handheld at 1/125s at f/11. Viewing it in 1:1 in Lightroom at is tacksharp with very fine detail.
If you print at large sizes (40x60" and 32x96" being my most common sizes), you can see a difference. At 22MP, out-of-focus areas, and areas which are slightly motion-blurred, are sometimes not apparent, as they are lost in the general mushiness of the image. At 42MP, though, and especially with even higher-resolution images stitched from multiple frames or from a MFDB, these flaws are quite apparent - instead of everything being equally sharp, you can see that one plane is quite visibly sharper than the rest of the image (making the rest of the image look 'wrong') or you can see motion blur where it wasn't previously apparent - stars may appear as short streaks, 2-3 pixels long, whereas a lower-resolution image would just render it as a single-pixel blob.
Basically, higher resolution gives you a better-quality print - razor-sharp 40x60s, for example - but this better quality also reveals every little technical flaw that isn't apparent in the mushiness of a lower-resolution image. You can hide the flaws by downsampling back to a lower resolution, but then you're back to the whole image looking mushy and no better than if you had taken it using a lower-resolution sensor.
