Transparency is key. The way we do it at ACM TOCE is having a public reviewer guidelines guide, and then when an editor things that a review deviates from it, they point to the specific guideline. Guidelines are community-sourced and approved by the editorial board. Process wise, that all works; the problem is that most Associate Editors have difficulty enforcing it, because reviewers say "if you don't like that my review calls the authors idiots, then you don't get my review." Reviewers have all the power, and this would shift some of it to editorial boards to ensure some agreed upon notion of quality.