i may have spent six days thinking "reagents" referred to the total number of inputs pulled, rather than the number of input glyphs placed. this solution is the final result of that misconception, using eight inputs and solving in 20 cycles for a "primary" score of 160.

interestingly, there are two very different recipes that lead to a low score in this new metric. if we use five mors inputs and four fire inputs, the theoretical cycles limit is 17, resulting in a score of 153. after a lot (really a lot) of effort i made some very promising half-solutions but i couldn't quite get the pipelines to unite properly. although the latency needs to be near-perfect at almost every stage i think 17 cycles is possible.

the other recipe uses six mors inputs and two fire inputs. 20 cycles is the lowest possible cycle count in my estimation, which this solution achieves. my solution for the real metric was thankfully only a minor modification of this solution.