Minimal difference in sharpness/CA ?
Correct, difference is minimal given everything I've seen thus far - unless you make a living shooting flat sharpness test charts.
The 50 f/1.2L was designed with uncorrected field curvature+spherical aberration with a priority on bokeh.
If you shoot flat test charts, that could be a problem. But since things we shoot generally are not a flat test chart and do have depth, real world performance in the shots I have seen appear to have minimal difference sharpness/CA. Wide open the Sigma does have a slight advantage in sharpness/CA real world use, but I expected that given the larger f/1.4 retrofocal design - at narrower apertures the Canon actually appears to pull ahead of the Sigma in sharpness (likely due to reduction of field curvature).
Field curvature does not make a lens subpar even though it will not perform as well on a test chart (just ask the $10k+ Leica Noctilux). In the end, the real world performance is what counts. And, some lenses will sacrifice some test chart sharpness/aberrations for superior bokeh and real world performance.
This lens is really in an entirely different class than Canon 50/1,2
At first, after only viewing the Sigma vs Zeiss, I was going to say that the Zeiss is clearly better through at least f4 (that's as far as I looked). The Sigma simply puts the 50L to shame, by a large margin. If you leave the Sigma at f/1.4, it still remains sharper than the 50L until about f/5.6, which is astounding.
You have to be really blind if you do not see the differences
The Zeiss 50 f2.0 was the best 50 for Canon by a larger margin, before the Otus, which technically isn't a 50, but oh well. The Zeiss 50 f1.4 however is absolute rubbish. So if the Sigma is worse than that it would be demolished by the Otus in the same way the sf-s 55-250 gets destroyed by the 70-200 mk2.
I can't see how the Sigma can be only slightly better than the 50 L and still compete with the Otus, that just doesn't add up. I'm beginning to wonder if the 50 Art lenses tested are all preproduction with high copy variation when the results differentiate so much.. But in some cases the example shots are just done wrong.
I downloaded the Zip with the raws from the 50 Art and I wasn't impressed at all, and with the first images shot by the Otus they blew my mind, so either Sigma lied, plain and simple, or people can't shoot or it's very high degree of copy variation and or preproduction lenses. This also of course mean I have to buy one, and buy it new so I can exchange it 4-5 times until a good copy is found.