82R, and 82 hex loop (design mandates that they're the same).

First thing that hinted me that P3 is possible is the configuration to wand a two-part brick to form a monomer, and then hammer the thermals to properly fuse it (break salt-salt bond and push left side to bond again).

The next thing is logistics with wands. I can't imagine a wand initialization mechanism within 4H, but stumbled into this idea of using another wand to make previous arm lose its wand. (If you see earlier iteration titled expelliarmus, this is why.) Initialization takes form of first monomer doesn't have this extra wand atom.

Since every monomer past the first one needs the wand atom for hammering to not explode, the arms need to be spaced exactly so that each wand is transferred to the next brick (instead of second next brick and so on). Therefore, the brick shape at wand-transfer time determines the amount of work to transform it to end form (which isn't compatible with wand-transfer action), and that in turn determines the speed of the machine.

It's neat that when I found the two height wand-transfer shape here I could start forming the right triplex bonds right out of the gate, and several transform steps can happen on the same glyph.