2*12+3+9*6+2=83

I feel pretty good about this iteration of my design. No extra delay comes from making the giant chain to get full throughput for the first 12 inputs- The bottleneck is mostly just the actual construction of the final molecule.  Improving that seems difficult because the period for bringing new earth/air bends to the bonder is 9 cycles & the period of my trackloop for bringing water/salt bends to the bonder is also 9 cycles.

The water/salt bends are limited because the bonding happens on top of the input.

The air/earth bends are limited because the arms have to make a rotation & pivot to avoid the calcifier from the same access point.  If part of the movement to the bonder involved a translation, there might be potential for decreasing the period of that section.

So essentially, my other bottleneck is the long path which the input has to take to reach the duplication glyphs & come back to the bonder.  That's where the +3 is from, then the +2 is from having to take a specific path to the output to keep the height at 4

Last minute gold shaves aaaaaaaaaa