Comment 275 for bug 85488

The following seem somewhat relevant:

http://lkml.org/lkml/2006/9/14/252
http://lkml.org/lkml/2007/1/14/98
http://lkml.org/lkml/2007/1/15/92

Someone seems to have known about this even in January, but there doesn't seem to be a good way of fixing this, apart from perhaps having a separate kernel if USB_SUSPEND must be enabled somewhere.

(Apparently the scanner is reconnecting and getting a new USB ID after an automatic suspend/resume, and as such is being non-compliant with the USB specifications... or something like that. The suspend occurs after two seconds of non-use.)

The workaround some people have used above is to install a daemon that keeps the scanner awake by polling it -- this prevents the scanner from going to sleep, and thus no suspend occurs, therefore there is no problem for resuming. That's not really a fix, though... IMO it's more like a kludge (which is probably worse than a blacklist).