The problem I've found with Windows 8 is that there is no visual difference between clickable text/icons and nonclickable text icons. The whole experience comes one of trial and error where you tap on random things like an idiot to see what elicits a response and what doesn't.
The web's way of solving this problem is by underlining text that can be clicked. HN violates this in many places, but at the very least text links underline when moused over - not ideal, but works, and is something that doesn't exist in touch-land.
Ditto, we've been using rollover effects for years to demarcate clickable icons from the unclickable. This entire strategy is now kaput thanks to touch.
We don't necessarily have to have drop shadows, but what we do need is a consistent and unambiguous language that clues the user into how the UI operates. Windows 8 currently lacks this in a really, really bad way. We can do this without skeumorphs, but we still need to do it.
And therein lies the greatest problem with this flat design trend - many designers are going flat without remembering that skeumorphs carry a lot of user hinting that you have to replicate.
I remember initially HN didn't strike as a particularly intuitive site. It worked and I am using it because of the community and the content, I am used to it by now.
Ok maybe there some value in having users slowly discover cool features by randomly trying to click things. I don't know. But at least it, seems, the core functionality should be somewhat obvious.
Buttons on the web for every link don't work, but I like some special markup to show links (traditionally it is the underlining but anything else can work as well).
To me there is often a tension between making something looking cool and making it functional. I'll admit the look of the flat interfaces maybe cleaner (whatever that means...) but unless it is meant to just be looked at, it also has to be functional.