The other solve has a big loop in the center where partially transformed reagents move on bottom and center and empty arms go back on top. It's the first time I can see the throughput at each section - 18R at pulling, 24R at assembly, some insignificant R at final touch up. Despite lower R I didn't build it out because sending catalysts to the touch-up side makes it hardly less cycles than this solve.

This solve has less regular center - each reagent squeezes past arms differently. Debonding the third reagent was thought to be a pain but turns out being useful work to get it into right rotation.