I've played with the SQL 2008 hierarchyID data type a little bit, but only a few minutes, and I've never used it in a production environment.

I have tested pre-2008 path-hierarchies, and found them to have all the negatives of denormalized data, and few of the positives. They are slower to update or query than adjacency, much slower to query than nested sets, and only slightly faster to update than nested sets. But that's basically because it's all string or XML functions. I'm not sure how performant the 2008 version is.

I'll see about getting a dev copy of 2008 this weekend (after payday), and see what I can do by way of speed tests.