This puzzle made me sweat something fierce. I have understood the *concept* of prebuilding for a while, but have consistently managed to avoid the hassle of actually *implementing* it. Here, we have the first case where a puzzle has forced the issue in order to hit min.  What's more, L=1 on all four inputs is a punishing limitation on time and geometry both!  And my Week 1 success notwithstanding, Latency remains a challenging metric for me.

But after four tense days, I did finally find this solution.  I expect it will be entirely uncompetitive on secondaries, unfortunately, but these myriad fat 3-swings were literally the only method I found to allow fire and water to "switch sides" of the machine without colliding with the ongoing animismus process.  Every more space-conscious layout I tried ended in failure.  There is one saving grace, however: I never need to swing a completed product.

**********

I am very curious to see how difficult the broader community found this puzzle.  Will we have only ~10 submissions at 20c, or more like 20+? I find it impossible to judge based solely on my own experience.  I did struggle, but when I step back, the parameters here do not seem like they should be as difficult as I found them to be.

Either way, I am satisfied that I have achieved the hard thing--min cycles on a required prebuild with zero spare latency to play with.  And I won't be so hesitant to try prebuilding anymore, after this!