That totally explains it. That was one side of the output bonded right to the case ground. So when your table was grounded via the ground wire on your grinder or other power tool that has a grounded case, you had a nice big ground loop. I am somewhat surprised that the machine could not detect that it had an internal short to ground on one of the outputs. Out of curiosity, was the short on the bus that fed the torch lead or the work clamp? There really should be more of an air gap between something like that and the case. I know it's pretty tight in there, but is there a chance that something got bent, and there is room to bend it back away from the case? Relying on a thin piece of tape or even heatshrink in direct contact with the case is not good. Especially if there can be HF on that.