Do big taxonomies ever work?

Ever get the feeling you’re watching a failure in slow motion? A project at my current employer – not one I’m directly involved with, let me say – is building a major new government website. They have pre-defined an exhaustive taxonomy which will drive the whole site; content authors will tag everything they write with an appropriate subject category, and the system will place it appropriately.
The tags may or may not correspond with the website’s front-end presentation structure. The explicit idea is to strip them of ownership of ‘content silos’, and theoretically deliver a more joined-up experience. My fingers are crossed for them. But…
I’ve tried to do similar things myself in the past; and I’ve come to the conclusion that big taxonomies in big bureaucracies simply don’t work. People don’t generally care enough to do it properly; it soon descends into a game of ‘how little can I get away with?’ (Feel free to tell me I’m wrong, by the way.)
So why do ‘tagging’ applications like Flickr or work? Mainly, I think, because your primary audience is yourself. It’s in your own interests to tag things properly, or you won’t find them again. If you’re tagging for Technorati, you know good tags will bring good traffic, which is the main reward mechanism for many (most?) bloggers. You are your own motivation. And yes, there are some truly customer-centric public servants who will go the extra mile. But most won’t.
Actually, I don’t entirely accept that tagging does ‘work’: not absolutely reliably. A Flickr search won’t necessarily reveal the particular photo you need, even if it’s in there. A quick click round reveals very different approaches to tagging, and link descriptions in particular. But we apply a very different quality standard. You might come across something useful, you might not. And if you don’t, it isn’t the end of the world. But if you’re looking for a key government document, and the tagging don’t surface it, you’ll be furious – and justifiably so.