I've always been very reluctant to use the CLR. On a completely unrelated tangent recently, I begrudgingly threw CLR into the ring of a string splitting performance test (blog post forthcoming). Actually, the reason I was performing the test at all was because of a couple of blog posts by Brad Schulz, where he shows some great enhancements to typical XML splitting mechanisms. (Articles are here and here.)
The results were astounding to me. Again, I will provide far more details later, but I was convinced that the mere overhead of the CLR would make it slower than other solutions, particularly on smaller strings. (Think about how hard it is to shave 5 minutes off of a 20 minute drive.) I was surprised that this was not the case at all, and as a result, I will be testing various ways to implement the CLR for various string functionality, such as splitting strings, and of course validating e-mail addresses and other RegEx necessities. I will also be interested to test it for file system operations, such as browsing a directory list, and queuing files up for BULK INSERT / archive / destruction.
So, CLR, you haven't yet found a place in my heart, but you have found a place in my databases. And it is due in large part to Adam Machanic, whose post last April on splitting strings has become one of my new favorites.
I hope this will be enough motivation for all of you other holdouts out there to give the CLR a shot!