This post is motivated by recent discussion (1, 2) concerning the relevance of anonymous peer-review in specialized areas of physics. From general perspective (which I always recommend to consider at the first place) anonymous collectivistic approach leads to the lost of personal motivation (which lead to the fall of communism, BTW) and it slows down generation of new ideas. While too individualistic approach fragments science and it slows down acceptation of new ideas.
In addition, there is always bias given by fact, if we choose reviewer, whose scope of interests overlaps with scope of interest of author, it becomes biased due the possible conflict of interest or existence of personal coalition.
If we separate the scope of interests, we increase risk of incompetence of reviewer. This risk is the more pronounced, the more science becomes specialized - which effectively means, above some critical density of information peer-review process isn't effective anymore.
Now we are dealing with two dimensional matrix handling distance of scope of interest and anonymity of peer-review process. The possible solution is to add time dimension into matrix and to make whole process as transparent, as possible ex post. In my opinion the most effective approach would be to keep peer-review as blind, as possible. BUT after publishing of article, it's peer-review should become available together with names of reviewers.
Of course, here's an apparent limit in density of information again and from long term perspective, every source of information should be published with minimal delay despite the result of peer review.