i heard about google-code-prettifier last year, but today i took few hours to evaluate it for my needs. The result is here:
• Google-code-prettify Examples
it seems a generic heuristic based parsers have a lot ways to go. Basically, i find that the prettifier has a lot problems when dealing with code that processes url, html, xml, etc. Basically, if your source code contains a lot regex, url, ol html code as string, it'll confuse the prettifier.
this may not be surpsing maybe because it is generic. Anyway, hopefully this is some useful info. I hope the prettifier can improve on this a bit... i don't know how though, but basically when your code contains complex regex or raw html as string (often needed in web apps e.g. perl, python, php, js), it botches the page.
PS the read-me page still points to the old version.