This is a small diff but we had an absolutely mindwarping time debugging this issue with Maya and Charlotte. One of the things we found out is that somehow, the firmware times out responding to a Get Status request (handled unconditionally in libfx2 by responding with two constant bytes), but only on Maya's Windows laptop; if I take the same series of requests, issued from my Linux laptop, and issue them byte for byte with roughly the same timings, they succeed each time.

Completely incomprehensible failure. We eventually gave up and drilled down the issue to this root cause, which actually doesn't quite makes this 100% successful, but makes it so that using the ETW USB tracing on Windows eliminates the low-probability failure entirely, which is the point where you just give up on debugging and ship the goddamn thing already.