The parts are still trickling in, but I'm waiting for the rest of the PCBs to start the mass-production of the workshop kits. Hopefully they will arrive tomorrow or Friday, so I can do it over the weekend.
Less good news is that the tracking number for the 200 servos that I ordered a week ago is not working (the seller gave me 4 different ones so far, none working). It's too late to order this from a different seller, and there is no brick-and-mortar shop that I know that has those. So it's possible that I will have only 10 complete kits, and 15 kits without legs. I'm still hoping that the servos will arrive on time, but there is nothing I can do with this. Frustrating.
Finally, on the software side, I made some progress debugging the I²C issues (after two evenings of analyzing the scope). I even found and tried a completely new slave library by Pololu, but decided to switch back to Wire.h for now (I might use that library later, especially since it has a workaround for the I²C bug in Raspberry Pi). Anyways, the issue seems to be that the Arduino slave is sometimes too slow for the 100kHz I²C, and the ESP8266 implementation doesn't support clock stretching, so it can't slow down when needed. The good part is that I found a way to compile a version of the firmware with slower I²C, so this shouldn't be a problem.