Downvote for this. Questioning or criticizing the author's credentials is a poor substitute for questioning or criticizing his thesis.
The author is posting a list of his personal gmail bugbears. We all understand the author doesn't care that you can see the email in the reply window, he'd rather a much bigger space to write.
I think we can safely say that Google test UX changes with actual users. Asking the parent for some data to back up that most users don't like the change is reasonable.
Also making the compose full screen would result in a very wide compose area, which is widely though to slow down reading (eg, of what the gmail users has just written).
I've actually flagged the submission: evaluating UX based on opinionated users without data is simply poor.
Slight trolling aside, I think this is a fairly obvious example of how a purely numbers & statistic based design process can go off the rails.
But in reality the people whose promotions and bonuses are riding on a successful launch of the redesign are also the ones cherry-picking the metrics of success. It doesn't matter if it's universally hated and actively drives people to use IMAP clients. There will be some bullshit statistic that'll go up. Hell, most metrics are going up anyway for a service like GMail, and it's not like this is being A/B tested.