The frame for the machine is a parent component that hosts a few 3D printed parts: the gantry, the two gantry legs, the base, and the backplate, which connects the base to the gantry legs. This still needs a bit of work, but the general gist is here, with only half of the legs and backplate shown, since they mirror. Also, updated colors to reflect the stash of PLA I'm actually using to print the parts out, with each color representing a rigid collection of stuff:

This links the kinematic loop of the machine together, from cutting tip to PCB blank!

Drawing it out like this helps set my mind at ease; so far, I've tested everything but the bottom third or so, where the force lines go through the Y-axis. Or put another way, I've stacked three of four axes and run a hand-actuated cut (granted, in only one direction) and haven't had any issues with stiffness from what I could see.

As part of this design, I added flexural overconstraint relief to the lead screw mounts. These parallel beam flexures will allow the bearing to translate left and right a bit as needed, while still keeping up some degree of axial stiffness. This is orthogonal to the overconstraint relief flexures built in to the lead screw nut mounts, so hopefully this prevents lead screw binding (noting that it wasn't an issue on the test gantry leg, which had a rigid mount, so maybe this is yak-shaving...):

Seems uh pretty good. The parts bolt together so I can print flexures and linear guide supports in preferred orientations, and to generally make assembly and printing and the rest a bit easier. Still to do, double-checking clearances, scootching the assembly around and fixing offsets (turns out designing complex assemblies is hard), adding feet, adding a front base support, adding motor mounts, adding limit switch mounts, etc. My hope is to print and test a full etch-a-sketch setup this weekend: carving a PCB by hand-rotating lead screws! If this works, I'll design pulleys, order belts, and generally push towards a powered test. I bet tramming will be a bit of a thing; maybe I'll try sweeping the substrate with a dial indicator.