You could do this:
1. Put each question set in its own subsection.
2. Make all those subsections of the same grading type.
3. Drop all but one of the subsections. (For instance, if you have 3 question sets, set the grading scheme to drop 2 of them.)
That way whichever one the student has the highest score on will count, and the other two won't.
It's not exactly what you asked for, but it's the closest thing I can think of right now that isn't really bizarre.
If you want the bizarre option: write each problem so that it randomly creates its text and answers using a Python script, then set randomization to "On Reset". That way when students reset the problem, they'll get a completely different problem. This is probably not the best option, especially if you want to make your data analysis easier later on.