This is the most interesting cost puzzle I've done! Solving within 110G required a lot of planning, though I never doubted it was possible. I suspect some people may get hung up on assuming a rhombus-shaped track, which is quite a lot harder.

My first solution was 281 instructions and 5 times the area, with a slightly different layout (the input dimer was aligned with the bonder instead of half-access). It was written in a single pass: I built a physical model of the target shape using hexagonal toys and moved it around by hand to keep track of the conditional motions, along with extensive notes on paper. I later spent a few hours optimizing it until I got here and saw nothing further to tweak.

I think this design could get another 10, maybe 20 instructions lower, and below that would require large changes to layout or construction order. I contemplated 1/4p, which would save about 30 instructions on dispersion, but probably cost much more in conditionals and sanity.

I don't think anyone who reaches 110G will 1p it, so it should be a reasonably tight secondary race.