I've found myself in classes where a group had exams, problem sets and other information from previous years (and solved) and they hid and refused to share those.
It was quite productive to go over all these exams as a group and strive to understand the questions, and particularly to understand the patterns, these were the concepts that lay behind the questions and once you had those you could answer many questions.
These days it isn't hard to find example problems and exams on the internet for your class or similar ones.
So while I'd never say that repeating questions is a good thing, I can't be too hard on folks who do (intentionally or not).
I quite liked the article. It'd be interesting to see if the benefits persisted over multiple groups though. I tend to see engagement increase when rules are changed and students need to figure out how to address the problem in a new way, regardless of whether they are collaborating or not. (anecdotal evidence there only sorry ;P)
If exams are more problems than exercises, and you are able to solve most of them before the exam, then you've probably learned enough and the exam is mostly unnecessary.
Needless to say after I recognised the pattern from the first exam I got perfect scores for the rest of the semester.