I've served on the program committee of the main conference in my field. It has double-blind review. In about 60% of the good quality papers I knew the person or team doing the work. In another 30% of the cases doing a literature search for other work pertaining to their sub-field showed who the team/person was. Most of the papers that were totally unknown to me were not very good. So I'm not sure how well double-blind review actually works for scientific specialities. Someone who is expert enough to properly evaluate a piece of research is likely to recognize the research focus of most significant groups in their field.

When I review a double-blind paper, I consider it part of my duty as a reviewer to avoid figuring out who wrote the paper. Sometimes I will accidentally figure it out when I look at related papers, in order to better understand the contribution. But even when I realize "Oh, right, it's probably the same group," there's still some unknowns in who, exactly, the authors are, and their order. (In my area, mostly computer science systems, author order roughly maps to amount of contribution.)

I'm usually a sub-reviewer, not usually on the program committee. And some of the conferences I review for are general enough that I may get papers in my general area, but not my current focus. So it may not make as much of a difference for the program committee of a tightly focused conference, but it may for all other cases.

Two issues. (1) Nature editors desk reject most papers without review, so editors should be blinded to author identity as well for this to sort of work. (2) By still allowing single-blind review, it gives the reviewers the notion that the authors want to be anonymous because they are not famous scientists, because a renown scientist with lots of Science/Nature papers would probably choose single-blind.

As sago suggested, even if the review is double-blind, it's often very easy to figure out the authors: just read the introduction and look at what papers it references. The introduction will cite a lot of existing papers that the authors/research group have published.

I'm concerned that it can't be truly double-blind. Any high-profile work can become well-known within the rather small academic community, and a reviewer may know the identities of the author(s) regardless.

Going through the motions and pretending that this is truly double-blind may introduce even more peculiar biases than existed if they had merely chosen to make it single-blind.

I image reviewers would typically know who the author is, or be able to discover it quickly, if they are in fact familiar enough with the field to meaningfully review the work. Similarly, authors can often guess who reviewers are based on the nature of the review. I don't think a double-blind strategy really addresses the problems with modern peer review.