The solution is entirely 1P except for arm 15, which requires initialization. I didn't put enough time into this puzzle, as shown by me cutting almost 20 cycles on the final day, so I hope this doesn't do too badly. I hope someone submitted a solution that pulls a bunch of reagents at the start at max rate then processes them in multiple places. I decided trying it was too much work, but it would be interesting to see if someone got that idea to work.