Not sure either, a dumb way to debug is to use evdev on the all the 
entries and search for the source of your events in old kernel. Ideally 
it should be available on by-id. But if it is not, it will be a little 
bit more complicated since you'll have to search on every entry of 
/dev/input/. When found, try to look in the need kernel if such entry 
also exist and have the same behaviour (technically it should not but 
just too be sure). When found, it should starts too isolate the issue 
somehow to a kind of device.

Also try to get the hid report descriptors (there might be more than 
one) from /sys/bus/hid/devices/*/report_desriptor for your device. It 
should helps maintainers to see if it is an interpretation error.