So, I originally used two arms for my cost-leaning initial solve. Then I tried a singular piston, and vaguely optimized that all night on sunday.
Then, when I couldn't sleep, I wondered if one arm on some track could do. A 1-length arm on 4 track has 12 spaces around it (5+2+2+3) which covers the 11 needed to solve this puzzle [2 to cut, 2 to bond, 2 to project, 1 for salt, and 4 more to grab inputs and outputs], but costs the same. A 1-length on 3-track is not enough, but a 2-length on 3-track is. Can we do better? In theory, yes! A 2-length on 2-track has 12 positions it can reach, though it might be tight.
That morning, I realized that yes, that would be *very* tight, and lengthened the arm.
I wonder whether Critelli cared for efficiency and optimization, or just results. I feel like I'm walking in the footsteps of people who walk in footsteps. His are long trod-over already.