A fast ultrawide lens that is dead sharp in the corners at 50mp doesn't exist.
Your best bets will be the Otus 28mm f/1.4 ( not really wide ) or the Leica Summilux 21mm f/1.4 ( on Sony Alpha ).
Even then - these lenses will still have lots of vignetting at f/1.4.
Alternatively - go down with your megapixel requirements.
Star shots are all about wide apertures and limited time exposures (20 secs or less without a astro-mount), so the Sigma is going to be seriously looked at by astro-shooters....corners and all!
My Samyang was back-focusing so much, I thought I was getting pre "big-bang" exposures......black frames.....it was clear to me that I was back focusing beyond the formation of the universe.....big league stuff huh?
I spent many hours daydreaming about my breakthrough and was already preparing for the scientific world's reception of my "White papers".......explaining this time/distance fabric situation with all the appropriate calculations of how to find the exact plane of focus in order to capture the actual glorious moment of the"Big Bang" itself .....
Sat back,eyes closed, gin+tonic in hand, bathing in my newly discovered eureka moment, the world was finally my oyster,.... then.... I looked up to notice that I had forgotten to remove the lens-cap....
It was at this point I decided to write a thesis on over-active neurons and the lifetime of confusion they had caused me!
At high resolution there are already visible startracks at 20 sec.
It's the question if a star tracker and a less fast lens (but well corrected) might be the better idea for static milky way shots. In your otherwise great picture the corners are also notshowing pointy stars but some stripes. For example, an Irix 15/2.4 has an adjustable infinity hard-stop, costs less than 1000 bucks, so a bit of money is left for getting a star tracker.
The question is also, if - except some star tracking experts and lens nerds - will complain about these things if the see a beautifully composed picture with a milky way no on usually sees with bare eyes at our light polluted night skies?
I think, the Sigma is meant for more than only milky way shots - it's sort of good for these kind of stuff, but less good than specialized gear. However, that specialized gear doesn't help much with wide angle wide open shots. The bokeh is very neat and I really love the lens for this kind of things. If I were "only "in milky way shots, I'd look for something else.
You asked what subject is on a flat plane to the corners.....astro photography is just that!
I think you will find the most of the star deformations are caused by the maximum exposure of around 20 seconds, 10-15 secs is the point at which stars start to make trails ....which are always worse in the corners.
I just got a nice parcel back from Sigma Switzerland.
Last week on Monday I sent them 4 lenses to check - their Swiss guarantee includes one free check per year. For some it was the first time to make the travel. One got the front ring replaced (for free), the others were checked and the 14/1.8 needs to go again to replace broken and glued lens hood against a new one coming from Japan - but I didn't want to leave the lens there for 3 weeks.
Now I'm again thinking about closing the 85/1.4 gap with a Sigma - or go the Tamron way this time. -_-