Given that only EV_KEY events are being mapped, without their scancode
counterparts, we just drop all scancode events to avoid having to map
them too, to syncing it with the corresponding mapped keys. This is
expected to be harmless.
Even though Peter Hutterer commented that it's not technically needed to
send SYN/delay when the sequence of events is of different keys, it
seems to be shown necessary to do so, otherwise events are dropped in
random situations.