Avenged:
Jobs were roughly split between atom identification and previous-atom-matching.  I worked mostly on the identification side.

I originally had a design doing something similar to this but turning a wheel for every purify result, but it occurred to me later, "do we even need the wheel if we are just doing that".  To which the answer was no. Quite happy with this setup from a cost and area standpoint.  Quicksilver is one of the main limiting factors for us going much lower than this.

We had a few other ideas for previous-atom-matching but none of them seemed as good as this, so this is SURELY the best solution possible tm.

Skyhawk:
This solve probably leans too far towards an area solve, even though cost seems to be more valuable for the sum. However with this arrangement, it seems like getting fewer arms in position to do more jobs feels like it would require more track than it's worth. 

For alternatives, I had the idea for doing identification by building sticks of 6 quicksilver and subtracting as many as needed to turn the atom into gold, but then I have no idea how you would check for matching length sticks.