It solves the puzzle. Arm 10 is a huge bottleneck, forcing rate 24 when all the other poorly optimized stuff is done in 20. There's definitely a better way to do this, but the set of possible approaches is so incomprehensible that I don't know where to start.