The fact Sotnikova got PCS that high, higher than Kostner's and almost as high as Kim's is absolutely ridiculous. She has poor posture, carriage and lines, both of her programs had very little relation to the music in terms of style or character and her timing was mostly off. Her basic skating is pretty good but it's not great and it's not at the level of some of the other top skaters either.
Her technical elements were extremely impressive in both programs and a few quibbles aside, her high TES scores - in relation to the other skaters at least - were largely justified. But the PCS were not.
As compelling as the 'Russians bribed everybody' discourse is, cultural, social and psychological processes are usually more complex than that. I've been thinking about the possible reasons for the result going this way and the possible factors I could think of were:
Sotnikova genuinely skated well technically in both programs and bolstered by the huge support of the audience, the judges simply went with her on the PCS as well. Anchoring, plain and simple. An overall positive impression got carried over to all of the marks.
Secondly, not all of the judges had ballet or dance training and even if they did as skaters, they might not know much about it in terms of looking from the outside. So if a skater is totally off-time and/or their posture is weak, they might not even notice that.
Thirdly, if you ever try to judge a program in real time, you will notice that when the program is over, you have spent pretty much the entire time thinking about the GOE of the technical elements. The music finished and you probably don't even remember much about what happened in between. Which is why panels should be split into GOE and PCS judges, IMO.
A system for marking PCS still doesn't exist and the criteria are pretty vague. And time is very limited. So given all of these constraints, it's understandable the judges simply go with their gut feeling and don't give PCS marks much thought. How? When?
Finally, the looming shadow of the corridor means you are likely to mark how you expect the others to and will try to keep all of your marks 'in line' without obvious outliers.