I wonder more what the self-selection for people who are willing to fill them out does to the results. Is it really accurate if you get, say, a 1% completion rate? Do those 1% represent the actual users?

One thing to consider is that some of the questions are open ended and ask about how your site could be improved or what users find most frustrating. Even with a handful of responses from a non-representative set of users you can get useful information to improve your site.