I’m certainly not qualified to answer that question, but I’ll give it a jab.
**Caveat- I have ad-blocker on chrome, so I am unaware if there is inappropriate content on some of the links I have here, but I believe these sites to be safe.
Camera makers are increasing megapixelage (if that’s a word) due to market pressure, and people’s real needs to print the size of a wall. As to market pressure, “this one goes to 11” is a perfect example of what marketing is. For professional event photographers, in a world where everyone thinks themselves to be a photographer, it could be helpful to have a camera that appears better than your client’s just to make the sale. As to real need, there are those who are paying 1000$ for multi-row panorama heads (http://www.reallyrightstuff.com/Shop/Panorama-Packages-for-Single-Multi-Row/Multi-Row-Pano-Elements-Package.html) just to get more detail in there shots. It would be cool for those people to just put an 11-24 (http://www.the-digital-picture.com/Reviews/Canon-EF-11-24mm-f-4L-USM-Lens.aspx) on a 5ds-r and get the whole thing in one snap. With a sharp lens, 50 mp full-frame allows cropping say, a full-body shot to a head shot. So that’s attractive to event photographers, and wildlife photographers. For wildlife photographers, the 5ds-r has a built-in crop mode, which should help the buffer. I don’t think that 50mp is overkill, frankly. In a few years we’ll be viewing our photos on 8K displays, which represents roughly 40 megapixels, so we’ll be grateful then that someone was shooting on a 5ds or a d810 back in 2015! 😉 Lastly, picture that is down-res’ed in PP to say 20 mp will appear sharper than a picture taken at 20 mp due to bayering.
As to too many pixels spoiling the image quality, The only thing we loose with more pixels is a per-pixel snr. Dynamic range and colour sensitivity are actually theoretically improved by having more pixels (by de-res’ing) Noise levels *per pixel* will be higher, yes, but by binning pixels, on-chip, or in PP, noise levels will be effectively reduced to levels on a similar-sized sensor with less pixels. 2 photosites receive as much light as 1 photosite that is twice the size. By having more pixels, there is less pixel area due to the circuitry (disregarding a bicmos sensor) but the micro-lenses make up for that, at least in theory. So basically if there’s enough photons to go around with all those pixels, a 50mp sensor will bring a phenomenal amount of detail to the table. If there’s not enough light, then we have to start reducing the size of the image – which will give it similar low-light results to a camera that started off with less (and larger) pixels in the first place.
A real world example: the 7d mark ii actually has a 40 megapixel aps-c sized sensor, with photosites binned together in pairs to create on-sensor phase detection autofocus. (dpaf) That means the pixels are twice as small as a similar 20mp crop camera, but the 7dii has a very similar final snr to any of its’ contemporaries – because the smaller pixels are combined to mitigate noise. The 7d mark ii also made a large step forward in terms of dark current and thermal noise, as well as the downstream electronics, so I would assume that such would help the smaller pixels survive until they’re binned. http://www.clarkvision.com/reviews/evaluation-canon-7dii/
As to iso levels, canon has claimed two things: firstly that they established the maximum iso based on what the professionals they interviewed needed. (that’s on a interview dpreview did with them lately http://www.dpreview.com/articles/5301008561/cp-2015-canon-interview-every-day-im-saying-speed-up) secondly canon claims to have increased the density of the cfa. (colour filter array, and sorry, can’t find a reference for that one, just know I read that a while back) That would give the sensor much greater colour accuracy, at the expense of a loss of light, which would result in lower snr. (signal-noise ratio) That would be a similar situation to the 5dii and the 1dsiii which have identical sensors. The 5dii is optimized for low-light (less dense cfa) and the 1ds is optimized for colour accuracy. http://www.dxomark.com/Cameras/Compare/Side-by-side/Canon-EOS-5D-Mark-II-versus-Canon-EOS-1Ds-Mark-III___483_436 Incidentally I find the 5ds is NOT the successor for the 5dIII (wait for the 5div) but is more a natural descendant of the 1ds series bodies which have always been about high megapixels (12 megapixels in 2002!) and controlled environments. The 5ds is aimed mainly at landscape and studio photographers who shoot almost exclusively at iso 100, and can control their own light.
I think Timtam has it: Canon has definitely set themselves up to make a jab at the medium-format world, with their lens releases in the last few years, and the 5ds-r, I think they have a good chance at it.
No, I don’t think we’ll ever have a “perfect” camera, because people’s needs vary too much – just look at the 1dx at 18 mp. Photojournalists want speed, and a 50 mp camera (roughly 60mb) at 14 fps would require 840mb/s write… While that will be possible in the future, energy and thermal restraints can’t allow that much power into a small camera body yet, let alone the cost of implementing such a technology. For now, canon for one has been increasingly segmenting their product lines for such reasons (think 7dii vs 6d)
I hope this is at least a bit coherent