I started my journey on Saturday (day of the release) with a straight forward, easy, 2 outputs in parallel 1 hour solve of 680g/303c/163a which I quickly improved to 610g/252c/147a.
I already knew that a finite solve would be faster, so afterwards I did implement a 1820g/171c/438a and improved and redesigned it until I got this 1595g/156c/375a, still on the first day. I could probably save some of the glyphs if I chose the correct rotations of the triplex bonders, but that was not really worth the effort. It also might make room to save a piece of track and thus a few cycles. But the last complete redesign only shaved 6 cycles and I don't want to optimize this monstrocity further. If someone took the time to do these optimizations, congrats on beating me.
I think IRG instead of ICG would have been the nicer metric, exactly to circumvent the finite solutions, but as it is they are sadly faster (or require 12 outputs instead). Same might apply to the B-side level, but I'm not bothering with a finite solve there.
A few days later I also implemented a solution that only does 1 output per loop and doesn't use any catalyst. I optimized its cost down to 430g/663c/123a, sacrificing a few cycles. It is my cheapest solution so far, but I never tried optimizing the 2 outputs/catalyst version for cost.