Time for my concluding rant.
I was not trying to prove anything and it was amusing to read how somebody found it amusing that I was trying to prove whatever, and another one quickly agreed. I did it for fun, to see if in perfect conditions it is worth using L primes, looking for evidence of that "3D look" (without bokeh), "popup colors", clarity, better contrast, and anything else of that sort but not for resolution differences. About half of the responders took it lightly but the other half took it as a signal for attack.
I did learn something from it, after pixel-peeping (the original resolution files are available now). A weakness of the 24-105 which I suspected but never bothered to check. It has less DOF (with pixel-peeper's small CoC) at f/11 than my 35L when focused close to the hyperfocal distance. The borders and the corners are not really softer (at 35mm). They just stay sharp over a smaller distance range. If you focus at infinity, no problem but then of course, you blur the foreground. This fact alone makes it worthwhile to use a prime for landscapes instead, when critical sharpness is important.
Why not testing it low light? What is the point - I know the answer. But as somebody mentioned in another thread - in some situations the IS is more useful that speed.