from biggie:
The solution name is because I quaff the tonic.

My impression of the cycles theory here was that the bent 3-atom salt-air-vitae was the clear winner.  It was the most convenient available input that could hit min cycles.  Vitae and two saltlikes was the only way to hit N=18.  Every way of attaching 3 connected atoms into this output takes 2 latency, so min is 39.  With the bent shape, you could do the final assembly on a single multibonder.

One input becomes two vitae, and two others become the left and right endcaps of the product.  But because of latency going from input through debonding, calcification, and animismus, the first output could never be on cycle 9.  So, we determine how many need to be batched together.

For the sake of cost, it is ideal to have a single animismus glyph operating with minimum 4 cycle repeat time.  This suggests that our best outcome is actually to batch the whole thing.  Batch of 2 comes up one cycle short at min latency, batch of 3 means having a tight route and taking an odd number of inputs before loop which is awkward for merging arms.  So I used the abundance of latency to try to eliminate any storage arms apart from the hex that pulls even inputs, and eliminate any track loops.  The result here is doing exactly what I think needs to be done to hit 39, and nothing more.

Started at 425, had iterations at 415, 405, 395, and then a clear jump to 345.  Several different versions of 345 exist but this one has the lowest area.

All iterations were shared in the Team 42 DM with morra and genius42.  Morra as a playtester is just watching along for enjoyment's sake, genius offered some testing and advice along the way.