Comment 25 for bug 430011

IIRC, initially the device powers exposing a profile that includes a USB cdrom (designed for storing MS Windows driver software not required for Linux). A kernel module then pokes the device into a mode that no longer shows the cdrom profile (and instead presents an additional serial port).

So perhaps it's a case of a serial driver implicitly hardcoded to presume that it's talking to a usb-serial endpoint, when in fact it might not be. I'm guessing that what we've had (working) for the last year is merely to blacklist and ignore the cdrom interface, whereas instead nothing *should bind* to it until it's been switched into another profile.