I don’t think we need to restrict ourselves to not repeat any of them. In fact it would be interesting to actively repeat some of the exercises to show how the same problem can be solved with different paradigms.
For instance, I really like Two Buckets and Tournament as object oriented exercises, but to also solve them in a language with immutable data or in a language with minimal data structures (bash) is fun.
I think we shouldn’t restrict ourselves unnecessarily, although I have a small preference for not repeating. But I could see some exercises being great for multiple months and help showcase how things can be solved in different tracks.