The limit was quite small. There was a relationship between the number of distinct primes you allowed to appear and the number of registers the Turing machine i

The limit was quite small. There was a relationship between the number of distinct primes you allowed to appear and the number of registers the Turing machine it modeled had, but I don't remember the details. I don't recall the choice of composites being very important. There was a pretty natural limit that was apparent to me at the time but I've long since lost my Fractran intuition.