I am reviewing a paper for a journal that has an open review policy. This means that when a paper is submitted to the journal, the paper is published online on the journal's website and when a decision is made, the reviews are published with the reviewers' names. If they want, reviewers can opt out and be anonymous.
I reviewed a paper where I was not anonymous. The decision was to request a major revision. I was invited to review a new version of the paper. I am still unsatisfied with some aspects of the paper but I do not think the problematic parts are sufficient to request a reject. However, I am reluctant to accept the paper with minor revision because I think the changes are substantial enough to require another round of review for checking the last version. If I could, I would propose another major revision, but the journal policy is that a major revision is either accepted (possibly with minor revision) or rejected.
So I thought I could contact the authors (whom I know) directly to tell them how to fix the issues, and then request a minor revision. Since they already know that I am a reviewer, this seems like it would not break any principles of good reviewing behaviour, would it?
Independantly of my precise case, and in order to make the answers useful to more people, it would be good to see opinions on what circumstances allow a reviewer to contact authors personally.
[BTW, I already asked the journal editors if this would be appropriate. At the time of writing the question, I'm waiting for their reply.]
Edit after journal editor answered my request:
The editor in chief replied by discouraging me from contacting the authors and provide suggestion how to deal with the situation within my review. There were several reasons evoked for not contacting the authors, but I will simply tell the one that I find most important: that it prevents documenting the review process properly. As a journal that prides itself of making open reviews, contacting the authors directly completely breaks transparency.