Comment 31 for bug 1579190

Tim, its essentially the same reason that we saw the problem manifest in some Wily and Xenial kernel versions, then seem to disappear in the next kernel version, only to return again in some future version:

The bug, which definitely existed all the way back through Precise, was that the driver could access the wrong data (an out of bounds array access) in some circumstances. The theory is that for any each individual kernel build, that particular wrong data it accesses might end up containing 'the 5 key' (making this bug manifest) or might end up containing a 'null' (resulting in no obvious manifestation for that kernel build). Its plausible that in Precise, the kernel builds just always ended up with a 'null' there.

The fix patch ([mainline] 50220de HID: core: prevent out-of-bound readings) prevents the driver from accessing that wrong data at all, which is certainly the correct behavior whether or not the 55555 problem was seen to manifest.