Other than the colors, I made a few changes to the original design, to balance the L2 frame and reduce friction on the drive gears.
The obvious one: the copper wire. I measured how much counterweight was needed to balance the L2 frame by putting a weight on a scale with a small hook on top, zeroing it, and then putting the hook on the L1 drive gear and lifting the L2 frame by the bearings until it was level - this read -15g on the scale, which meant I needed 15g of L2 counterweights. Since the plastic weight cages and screws were <3g, I knew I couldn't fit enough in (which McMaven noted in the build instructions). I had some 16AWG grounding copper laying around, calculated how much would be 15g, and wrapped the frame with it. Got a nearly perfect balance, or at least it doesn't move on its own from horizontal. Yes, it's kinda ugly, but I'll pretend it looks like a motor winding... (I balanced the L1 frame with 5 paperclips, but that's not visible)
Thanks to the new larger L2 counterweights, I was able to fill them enough to match the mass of the copper wire and replace it. I squished in a bunch of solder in wire form and there was some room left for more, so it appears to be a success. New pictures added.
For the L1 drive gear: I recessed the top about 0.3mm (printed 0.15mm layer height) for a circle just larger than the screws. This made the inner bearing post shoulder protrude 0.3mm higher than the gear and that way only the shoulder was touching the inner race of the bearing and no part of the gear could rub against the bearing outer race or the L2 frame.
For the L2 drive gear: I made the whole thing 0.4mm shorter (printed 0.2mm layer height), again so the L2 bearing post shoulder was only touching the bearing inner race and no part of the gear would rub against the base.
I measured a beat frequency of 3.42 beats per second as built, and got to a somewhat nicer 3.515 beats per second after removing two balance wheel screws. I figured that by driving the whole thing at the equivalent of 3.5 beats per second I could "force" a better timing by letting it slow a tiny bit when there was less drive torque. The gear ratios work out that one full stepper pinion revolution is 20 ticks (1 L2 frame revolution = 3 L1 frame revolutions = 100 ticks). Driving at 10.5 RPM on the stepper pinion gives 3.5 beats per second, so that's what I set in software and it worked great without any further tuning.
Video of it in motion: https://www.youtube.com/watch?v=08naIMTs3k0
During the replacement process for the new L2 counterweights I backdrove the stepper motor to get the frame rotated to where I could get to it, and apparently that motor doesn't like it? I had been able to drive it with no problems originally off a USB hub that also had my mouse and keyboard and a CF/SD card reader off a USB2 port on my desktop (assume <500mA). Now it doesn't sound quite right and stops after a few seconds of ticking, getting stuck, and requires a prod to the balance wheel to get going again, both on the same hub and with a dedicated 1.2A supply from a phone charger. I may take it apart and see what I can do, but it's a hassle to replace all 12 screws for the L3 frame, so that may wait for a while.