I never said i know for sure what's wrong.
I just pointed out that things like these can happen if grounding is off AND some other problem (defect or a second incorrect wiring) occurs.
I know of this to happen, because in my work i often see trouble with different potentials which get shorted because of some problem (can't always be identified) and the fact that the PC has a shared PE and GND.
To me that means there needs to be 2 faults, of which one is the shared PE/GND (but that is a fact you can't do too much about).
You could power your PC through a 1:1 transformer (can be found), or isolate the serial connection, which usually is a bit easier, as well as cheaper to do.
The GND connection of your USB cable is pretty massive, compared to the usually quite flimsy DuPont wires.
So that wire is the bottleneck.
Do try the suggested resistor (which was mentioned and therefore should have never left out anyway, unless you know exactly what you're doing), it'll be the new bottleneck.