if you live in a dry climate, and you have carpets(isolating you from earth potential), it could be YOU that is the source of the problem, if that is the case, you are statically charged and basically the case is serving as a path for you to discharge your static build up. That actually isn't good, as you can kill electronics shocking them like that, especially if the shock is passing through the switch and pcb before making it to the grounded case. The switch could care less, but the LED's(or led drivers for that matter) wouldn't like thousands of volts passing through it like that.
The keyboard should be grounded to the case by the screws holding the PCB to the case, and the ground reference for the case should be provided by the USB ground shield. Technically, there should be 0v between the negative provided by the USB cable(usually the black wire) and the ground shield, but that is ideal. The ground is supposed to soak any stray voltage caused by EMF noise, so in reality, you should read a small amount of potential difference between them. Either way, the ground reference SHOULD be referenced to real earth, unless impossible, like if you are using a laptop with a virtual ground. You can also get away with just resistively shorting together the negative and the ground, but the board should already be doing this.
Basically, if I had to guess, I would guess it was one of three things. Either your problem is a short, your problem is that the earth is oversaturated with far too much AC voltage on it, or that you are heavily statically charged, and you have the acquired superpower of killing electronics with thors lightning. Grab a multimeter and test your own potential to real earth, and then real earths potential to the case. That should make it clear. With how much electrical mass the KBD75 has, I HIGHLY doubt anything is saturating ground on the keyboards side.
For reference, I read 0.1v DC and 0.3v AC between myself(earth grounded) and my case(Grounded to my device).
Also, I had another keyboard that wigged out when I did ANYTHING which generated a lot of EMF noise anywhere on the same side of the house. The issue there was the USB cable. I eventually came to the conclusion that if enough potential difference found its way on the USB shielding, it would arc to the data pins and make it malfunction. Either way, changing the USB cable fixed the issue. However, the repeatability of your problem kind of makes me shy away from that being the culprit, but it is worth pointing out. Intermittent shorts can cause a whole host of weird behavior, and is a pain to diagnose, especially on SMD PCBs.