Whereas it is often considered unethical to manage as an editor, associate editor or guest editor a manuscript from a researcher with whom you have coauthored one or more papers, this does not typically hold true for peer-review.
In other words, it is considered acceptable by most editors for me to review a manuscript from one or more authors with whom I have coauthored one or more papers, and even authors with whom I am currently collaborating.
But is this truly ethical?
When you are being asked to peer-review a paper authored by somebody you have collaborated with (or are collaborating with), you should be aware that you are on a slippery slope.
The very least you should do is:
There are situations where it might be acceptable, and situations where it is not. It is certainly not acceptable when you have a clear interest in the outcome of the review process (think of reviewing a grant application, where you would be benefiting somehow). From the other hand, when you are (one of) the clear expert on a certain topic, and had only a superficial collaboration with the authors, it might be acceptable (in some narrow fields it might be kind of unavoidable).
I think you have to discuss this topic much more nuanced.
Sometimes, people become co-author on a paper where you also collaborated, without you actually knowing them very well or at all. In this case, another author on the paper (likely the principal author) knows both you and the other co-author, both of you did their part, and that's how you ended up being co-authors. Under such conditions, I would not object when you review a publication in which this other co-author is involved. You hardly know them, so you are reasonably objective in my opinion.
On the other hand, if you are talking about people with whom you are regularly collaborating, whom you regularly invite for co-authorship and they do the same with you, whom you know very well, and with whom you potentially have a collaboration going on right at that moment; then this is something completely different. I would never agree to review a paper where any of those people played a significant role in performing the study, and I would consider it unethical to do that, because you can hardly be objective anymore.
Having said that, I must admit that there is still a lot of leverage. You may have ended up being co-author on a paper, where you still have a very professional relationship with the principal author (i.e. they were not members of your work-group, supervisors, etc.) and are still reasonably objective, although you published with them in the past. Additionally, as Captain Emacs already pointed out, some fields are small. At some point you know basically everyone in that field, and may have engaged in collaborations with lots of them at some point. It is in noones interest to significantly reduce the pool of potential reviewers just to stick to some over-the-top rules. I normally decide for myself if I can be objective in regard to certain authors, and decline to review if I cannot (and inform the editor if I know any of the authors on a significant level). And I just hope, other reviewers are doing the same.
I will speak about the practice in my field, and leave my judgment to the end (below). I think that in some areas of applied mathematics it is quite common to referee papers written by one's former coauthors.
In one subfield that I work in, I very frequently review papers written by someone who I have previously coauthored a paper with. This is partly because I have co-authored papers with a large portion of the active researchers in this particular area (it is quite small). In most cases, the editor (who may also be a former co-author) is well aware of this relationship. I can often tell who has refereed my own papers because there is a small set of possible referees and I know them well. I find that the most insightful and critical reviews almost always come from people in this small circle, because they know the area so well and can much more readily spot deficiencies. Meanwhile, reviews from people who are not well acquainted with the area are often quite superficial; reports that ask for minor revisions or no revisions in the first round are typically from such people.
My sense is that this small community is willing and able to provide serious criticism among its own members. The fact that most of them have worked together in the past (and may again in the future) does not stand in the way of this. That claim might be harder to defend in any field that is "softer" than mathematics, since reviews will necessarily be more subjective.
I'd say it's unethical if one of the following apply:
If everyone was duly notified, including the readers somehow, and it can't be helped - which can often be the case in smaller fields of research - you're ok ethically in my book.
(*) - This point can be ignored if it's well known than just one or two prominent people are doing all the reviewing, in which case there's an inherent bias by the very setup - but readers can be expected to know about it.
It is for the editor or associate editor to make the final call once the situation has been explained.
The two cornerstones of the refereeing process (at least in my field) are anonymity and fairness/unbiasedness. Playing with either is asking for trouble. Always ask: what would happen if conflict (real or perceived) was somehow revealed, or discovered after the fact by the editorial team? The publisher or editor might have to take action that is not in the best interest of the referee or the authors.
If you think there could be a conflict of interest, immediately contact the associate editor who sent you the paper and your relation with the authors; continue only once you have the blessing of the editors. Even if there is no conflict, you will come out leaving a good impression. (I was once asked to referee one of my own submission: I declined. Obviously this was a clerical error, and shows such errors occur in assigning referees.)
Sorry if this is such a strong response.
(The only exception I know is to conference proceedings, where participants are often asked to referee papers by other participants, and when some form of conflict of interest might be inevitable.)
