The important part for elorant (and me) is that CSS is awful to work with. And everyone in this thread is proving it by pointing out that you should use some kind of translation layer instead of actually touching it.
Well, there's two things at play here. The complexity of the CSS model in the browser (plus all the browser-specific quirks) and the CSS syntax. Less only solves the CSS syntax issue.
My reading of this thread is that it was started based on the complexity of dealing with CSS in the browser rather than dealing with the CSS syntax.
CSS also has serious issues with its design (not just syntax) - it's a pretty awful markup language, because it was created apparently without reference to hundreds of years of graphic design and text layout and missed out some essential typographic and design concepts.
To pick a few examples, until recently you couldn't have grids, multiple columns of text (still pretty broken), embedded fonts, and the box model is unnecessarily complex, which led to loads of bugs in the browsers. That's easy to say in retrospect of course, and lots of the problems are now being fixed or worked around, but it's a pretty gnarly language, and many criticisms of it are not based only on ignorance but on substantial flaws in the original design. CSS is awful to work with, and is only slowly getting better.
The slowness is largely due to competing interests & implementations by those interests. I don't think blowing away CSS completely would be a good solution - just look at the skepticism by many over Google trying to implement Dart in Chrome, and whether it'd be successful in eliminating JavaScript.
It does suck to work with, but it is workable if you put in the time investment to learn it, just like any language. Nothing that justifies flat out incorrect statements/dramatic behavior.
My reading of this thread is that it was started based on the complexity of dealing with CSS in the browser rather than dealing with the CSS syntax.