Philip, it is well understood that, as you say, when resolution of sensors increases, weaknesses in lenses is revealed. Clearly a full frame sensor with just ten pixels would be undemanding whereas one with 100mp would be very fussy. At some point in between, and it might be at 16mp 24mp or 36mp or 50mp, the lens will reach the limits, at some part of its field of view, of whether it can supply enough resolved information to match the resolution of the sensor.
The corollary of this is that even if that point is reached at, say, 34mp, a file taken on a 36mp sensor and then down-resed to 24mp should be at least as good as a file taken with that lens on a 24mp sensor.
If, as I suspect is the case with the A7 and A7R, a same-aperture file taken with a lens such as the 35mm F2.8 on the A7R, when down-resed to 24mp looks worse at some points in the frame than the A7 file, then there is a specific issue with the higher mp sensor
which does not relate purely to the increase in pixel resolution. I suspect it relates to the arrangement of micro lenses.
So I think that you may have misunderstood my point: downresing the A7R file isn't an attempt at bolting the stable door - it is an exercise to determine whether the weakness of the 36mp file is caused by the sensor having out resolved the lens or whether there is some other factor.
If the downresed file looked as good as the A7 file, we would suspect that the problem was that the 36mp sensor had out-resolved the lens at some points in the field of view. But if, as seems the case, the A7R file downresed looks worse, we should suspect other factors such as the micro lenses.
Tim, would it not be that the capture is when the result is determined and the downsizing is shutting the gate after the horse bolted? In that you are downloading a compromised file with low contrast and/or mushy corners, incapable of remediation.
My CY Distagon 21mm does very good corners on the a7r, better than on the a99, but like the 21 and 24 SEMs (which sadly have excessive beam angles) it is exceptional.
I believe all lenses must work harder on the a7r, this has been a consistent pattern since day 1. The mighty small one rewards optical excellence but punishes optical weakness. Of course almost all wide angles tend to drop off more and it shows in the MTF data. Same problem with almost all normal lenses - the otherwise great 50 Makro-Planar has just 50% the fine detail performance at the corners as in the image centre.
Photozone found the FE35 to have a significant drop off in midframe and corners, tested on the a7r; and it shows up in the slrgear Blur Index chart as well - tested on the a7r also. You should also see much better image centre resolution from the a7r than the a7.
You can simply get away with more design weakness in the lens with less resolution on the sensor. As sensor resolution goes up the gap between image centre and corners gets wider and much more obvious. D700 owners are well placed!
35mm lenses are simply hard to design for fine corners. People can wait for the next RX1 with the 36Mp non-AA filter inside, it will break the trend as that lens is ready for anything.