In the original post, I mentioned that the differential between my external 5V supply and the USB power was < 100mV. Generally less than 25mV, every time I checked. I still bricked the USB ports of several Megas. I wish I had checked the temperature of the chips, but it took me a while to localize the problem, because they did not appear to fail at once. It was only when I tried to reload sketches that I realized I had a problem.