This hexarm dropoff was my very first idea for this many output spaces, but for a while it seemed to me that it's a not a good fit for this metric. Eventually I managed to integrate it nicely into the machine. I believe this is only 2 necessary cycles off the the fastest possible solve. The small hex arm releases one cycle late, so the other salt can fit through, and there is another cycle of waiting for the big hex arm to get in position for the final phase. I liked cycles X arms, it's kind of a distant cousin of sum.