There is quite a big and obvious con: Segwit is untested code, meaning it could open up new attack vectors if the devs aren't EXTREMELY THOROUGH in its design and deployment. They need to think like attackers, trying to figure out all the possible attack or disruption vectors and then make sure it all works ok.
Not sure whether shill or just uninformed/stupid. As of December 31, 2015 there has been a Segwit testnet called, segnet. There have been various versions of it and it is being thoroughly tested.
Guess what is untested: Hardforking to 2MB.
Aside from claims from 'personal' testing, it is pretty much untested indeed. 2 MB block size limit is unsafe, this is why Gavin added sigops limitations to it.
There aren't any realistic SegWit cons I could think off... no one of the cons listed seem like a real con to me. You don't lose anything of value with SegWit, all you gain is better capacity without the hassle of a hardfork, what's to complain about really?
Because there aren't any. The only thing that I saw was complaints about complexity. However, these are usually hyperbolic and coming from people with limited or no knowledge in related areas.
