A little bit of a different viewpoint: Most people have focused on the clustered vs nonclustered aspect of the question, but when I first read it, I thought the point was the enforcement. Of course a pk must be enforced, so I almost answered incorrectly. Thankfully I've learned that most of these questions are tricks, so I reread it...

Edit: I guess Oleg beet me to the punch on that one. Cheers!

As to the second question, Of course having the option for a nonclustered PK is a good thing!

So many people use a surrogate key (identity column) as their primary key, even when it doesn't make sense.

Clustering your table based on _accesses_ instead of arbitrary design decisions can be a good way to get some extra performance.