The average interview score from a panel of interviews correlates very highly with performance at Google, interviewees report high satisfaction, etc.
If those things changed, i'm sure Google would change it. But why would they change it otherwise?
Yeah, I struggle with that, too. I've never interviewed with them, much less worked there, so everything I know or think I know about their interviewing process and work culture is second- or third-hand, but it sounds pretty miserable from the outside looking in. Still, they seem to attract a lot of people who really love working there, and they're producing things like Google maps, GMail, self-driving cars, Android, Google Docs, and a search engine... I'd be irritated if I interviewed there and was rejected (although I'm sure I have much worse interview rejection stories from far less prestigious employers), but I'm not sure that would necessarily mean that they're doing it wrong.
Does it? looking at the pigsty that is the gmail redesign, it sure doesn't feel like they're hiring the brightest they could.
However, as of 2014 average interview score was very well correlated with job performance (I can't give you further exact details because they are still marked confidential and i can't find a paper or anything that has disclosed it).
Note very clearly the word average above ;) Individual interview scores are definitely not predictive of job performance.