Double-blind reviews kinda' sorta' work. I've watched several research communities address this question. Some of the cryptography conferences use double-blind reviewing. It means that no one should refer to "in our earlier work, ...." And it means that reviewers are sometimes mistaken in their assumptions about the authors. But political decisions are sometimes made.
In the structural complexity community, this idea is brought up once in a while and dismissed. The community is small enough and tight enough that people tend to know about papers (not all, but a significant number) before they are submitted, and tend to recognize writing styles. At best, blind reviewing would simply remind the reviewers that they are supposed to be ignoring personal opinions about the authors. I guess that community trusts its reviewers to do so without that form of reminder.
