Perhaps we need a study of comments in the different percentiles now vs a year ago vs 2 years ago. Filter out comments that don't tell you something new. How does the remainder look?
The comment system is usually very good but for things like the Google/China post it is very hard to come back a few hours later and see what's new because of how things are ordered, I'm sure there would have been a lot of valid points in that thread which didn't get upvotes.
Is that also related to the way comments are now weighted against time, as well as value? Every new comment is momentarily - or longer - placed at the top of the comments page, so something trite will live above a 32 point well-reasoned post for a period of time, depending on how old the 32 pointer happens to be.
When the new comments went straight to the pile of 1 point comments at the bottom, there was less incentive to contribute late in a discussion as your input could be overlooked; in fixing that problem, there's now an increased noise-to-signal ratio as low-value comments filter their way down, rather than build their way up.
Also, part of the issue with scaling is that there are some types of comments that are fine to post when the userbase is small, but that aren't ok to post as the userbase grows-- even if they add value in some way. So the problem isn't just coming from new users posting bad comments, it's coming from old users posting the same comments. So maybe some sort of radical experiment to try to refocus the discussions wouldn't be such a bad idea, at least to see how it goes.