Pro: Resolution is dynamite, though we only have 5D3 data to base that on. Vignetting seems pretty welled controlled for such a large aperture instrument.
Con: Size (of course) and AF. 1 second to travel from end to end of the scale (that's 3-4x slower than a proper ring USM prime).
The good news about the AF -- the center point never swings and misses at f/1.4:
"In order to check the autofocus performance in a more controllable manner we decided to test it on our resolution charts. We chose very demanding conditions, using the f/1.4, not the usual f/2.8 aperture, because we wanted to see whether the central point provides you a reliable performance even in the area close to the maximum relative aperture. It seems the Sigma fared well enough – it hit the target spot-on in half of cases, with the quality of photos not worse than 10% from the best shot. There were 28% of photos a bit worse but still completely acceptable (decrease of resolution ranging from 10 to 20%). In the next range, from 20-30% we placed about 22% of all photos. We didn’t find misses greater than 30% at all."
The bad news? Don't leave the center point -- here's how it did (presumably at the f/2.8 they usually test with):
"When it comes to accuracy, already while shooting sample shots we managed to spot one alarming thing. The autofocus performance is quite reliable when you use the central point and its closest areas but it worsens notably when you use the outermost points of autofocus. In that case you can experience distinct misses in even about 30% of shots."
Testing was done on a 5D3 -- absolutely no slouch to focus with.
Thanks to LT for being (to my knowledge) the best show in town on AF testing.