(Context: theoretical computer science.)
I just received a review request; I have previously reviewed three previous versions of the same submission (twice for conferences, once for a journal), recommending rejection each time. For one of the conferences, there was discussion between reviewers and PC members; all reviewers agreed that the paper was “strong reject”.
Between each iteration, the paper has been lightly revised, but never addressing two fundamental problems: the novel content is very insubstantial, and the write-up contains substantial self-plagiarism.
The author(s) have/has extensive publication records, containing some genuinely significant work, but also a very large amount of repetition, often to the extent of self-plagiarism.
My current response is: accept to do the review (first letting the editor know that I have reviewed earlier versions and recommended rejection); then see what significant changes have been made, and if the major problems haven't been addressed, re-reject with essentially the same criticisms as before.
A generous referee might reasonably judge the work substantial enough for a third-tier conference, and not notice the self-plagiarism; so I guess the paper will probably strike it lucky and get published sooner or later. Is there anything else I can or should be doing to forestall this? Or should I accept that bad papers do get published, and that the damage the authors are causing to their reputation with referees/editors along the way is enough censure?