How do they do it? Prune "By X" every time? Or is there a more sophisticated system?

This seems to indicate some sort of cron just fetches the <title> tag a little while after something's posted and updates accordingly. Matches with my experience - I've seen perfectly good titles reverted to nonsensical garbage no admin would've manually reverted to.

https://news.ycombinator.com/item?id=6572910

But the title of this page was noise. The title was embedded in the pdf.

I think it’s manually done. They also moved the date to the end.