Exactly one month to this very day I wrote the article “When should you use NOLOCK?” which explained exactly when the use of the READUNCOMMITTED isolation level was acceptable and when it wasn’t. I also revealed that it not only takes out a Schema Stability (Sch-S) lock on the object (which is documented in Books Online for the hint) but also took out a special Shared (S) Bulk Operation lock when used on heaps. This latter behaviour is not even mentioned in Books Online for the hint and as far as I know, has never been documented or mentioned anywhere else before.
I therefore asked Paul Randal to check out my aforementioned blog post to confirm that I wasn’t going mad, and thankfully he confirmed that I wasn’t and that it was in fact expected behaviour. He also thanked me for giving him the idea for his final myth for his “More DBA Mythbusters(DBA-316-S)” session which he performed yesterday in which he gave and explanation of these behaviours. So I not only encourage you to check out my post, but to get hold of the PASS Summit 2011 DVD set and have a watch!