With as much as has been tried, it most likely is a bad machine. The BM1387 ASICs are really difficult to reliably solder with the equipment I have. Pins are sub-quarter-millimeter dimensions which means virtually zero solder, and the comparably massive belly pads can cause lift. Took us about six months of trial and error to find a stencil profile, placement method and reflow oven speed/temp profile that worked satisfactorily, but it's not perfect. It's basically an exacerbation of the same issue we had with R606, where there'd be pad contact or, at best, an unreliable solder joint, while testing that then gets disconnected once conditions change after some burn-in by the customer, resulting in dropped chips.

I'll keep making R606 and testing as strenuously as I can before releasing, but unfortunately until we start using a different, better ASIC in a new version, this is going to keep happening to a percentage of pods.