I have historically struggled with cabinets.  I surmise that this is because my favored duct-tape-and-zip-ties method of solving doesn't work well without an infinite canvas.

But today is different.  Today, I think I have slain that dragon. 

**********

My first thought on this puzzle was that it was probably conduit-limited. That meant the inner loop should be a multiple of 3 rather than 2.  With 2 molten iron required to produce both vitae and silver, that suggested an ideal target of P6 for the inner loop, and a total output rate of 18.

I assumed assembly at that rate would be the biggest challenge, so I tested that first.  I copied the puzzle, changed the inputs to silver, quicksilver, and vitae, and started dumping them into the assembly chamber at top speed. In the end, it *almost* worked.  The best method I found needed to pause the machine for 1 cycle in order to make the final bond.  Still, 19R seemed close enough to perfect for production.  My pie-in-the-sky target minimum for the concept was thus set at 127 cycles.

But my first assumption was VERY wrong.  Assembly was not the problem.  More difficult by far was designing preparation chambers that could support a P6 inner loop.  The animismus side came together easily enough, but my metal-side designs were a series of agonizingly close failures.  It required a minimum of seven arms--which, when combined with the required glyphs, inputs, and conduits, meant that every single hex must be used, with tracks only allowed if they were fully occupied loops.  Time after time, I found myself either one hex short, or with the last free hex in a spot that I couldn't make useful.  I gave up for the moment, settled for P8, and posted a solve at 173c.

********

Later, looking at the comparatively spacious animismus chamber, I tried a new tactic: placing quicksilver on the animismus side, and feeding one atom back to the metal chamber via conduit phasing.  Which is when I started internally cursing Yosh for the chunky mors, because the three additional hexes it claims caused exactly the same problem!  Every design was either one hex short, or had the last free hex in a useless spot.  Foiled again!

********

In desperation, I finally argued to myself that the metal side didn't actually need seven ARMS, it needed seven GRABBERS.  Could a hexarm rescue the concept? I had originally ruled it out because swinging the molten iron instead of sliding it would burn an extra hex, and there were obviously none to spare. But with nothing else working, I gave it a shot.  

Scarcely ten minutes later, I was staring at what seemed a workable layout.  Surely it couldn't have been so simple. Surely there would be some unavoidable BS collision in programming that would ruin everything.

But NO.  It DID work.  The right chamber you see here rescued the dream.  173c immediately plummeted to 133c.  Huzzah!!!  All hail Hexarm 1!

*********

Resting happily on my laurels, and with days to spare to the deadline, I started idly poking around for cheaper construction methods.  To my great shock, the weird triangle-track approach you see here, though not saving any cost, is *just* fast enough to make the last product bond without requiring a pause.

I had found the holy grail of 18R, clocking in at 126c.

This is one cycle faster than my first high-ambition target, and--if my calculation is right--only 5 cycles higher than the theoretical minimum with animismus and metal promotion in separate chambers.  Dragon slain, indeed.