Whats the worst that can happen, people might go through it and learn all the solutions. Well, let them, that's the purpose of the course anyways. But the question bank cant so small that it does not explore the full diversity of problems. And no one is claiming that all questions will be from the question-bank, throw in a few off question-bank odd-balls each year.
But how could they analyze the submissions to figure out (even approximately) who cheated who did not ? Apart from trawling their email and phone calls and wire taps that is....:-) I suspect part of the "forensics" was a bluff.
I can only guess that there are a few problems in the set that historically have a low probability of being solved correctly. So whoever solved those can be marked suspicious. But a test will have only a few of those.
But it sure sucks to be in a course where the instructor is unaware of the problem that QB is available and you are unwilling to look up the QB. Particularly where the QB was particularly designed for the top percentile.