not beautiful, but my first solve that's legitimately better tarcles than my 1-arm solution. A miracle! Still has a bit of room left for optimization, but I don't think I'm ever getting a tarcles score better than 360 out of this machine - the hex arm can't make half-products any faster than once per 5, which sets a hard minimum cycles for any machine using it of 120c, or 10c/product. Needless to say, that I got so close to this limit in the first place is pleasing, but it does put a pretty strong ceiling. I think the only options are to drill down closer to 360 by cutting the excess cycle off the machine period somehow and then cutting more purely to save latency, or to go back to a two-arm and find a way to take no more than ~31c/pair of products.