I do not see how using two power sources should damage the arduino. The difference between a regulated 5v power supply and the USB port should be minimal and not harmful.
I have now encountered a 4'th mega r3 that is encountering usb communication difficulties - and I'm about 99.9% certain I have NEVER plugged this one into two power supplies - only the usb power. The trouble began with seeing 'garble' when the mega was reset, my sketch should print 'status' to the serial port upon reset and 'sometimes' it was garbled. Huh...Maybe just trash on the line I thought. But then, when I pushed the reset button it began not writing to the serial port at all sometimes. I'd have to close the arduino ide's comm window, unplug/replug the usb cable back in, reopen the comm window, and then upon reset could get it to speak on an intermittent basis. Sometimes it would stop talking after 5 resets, sometimes after 3 resets, one one round I got all the way to 24 resets before it stopped talking. At that point I moved from the macintosh over to the pc and found the exact same symptoms (ocasional garble, and dying after some low number of resets).At that point I grabbed a mega rev2, loaded the exact same software onto it, and it works perfectly.I went back to the mega rev3, reloaded the software - and it's intermittent, and it got worse the longer it was on. It looks to me like I'm catching the beginnings of a usb controller chip failure.So, it's beginning to look like a bad batch of mega rev 3, or a design flaw, one or the other, to me. I'll be working on the mega rev2 all day and tomorrow, and will keep it continuously powered up and will report back later.
Do you know where I can find instructions for reloading the 8u2's firmware? I looked for that info but failed to find it when the boards began going bad.
Regarding "Usually digital chips do not go intermittent (they can but usually dont)" - I understand and agree but that's what appears to be happening. But something weird is happening that's for sure. In the case where the boards were plugged into the two power supplies it doesn't make sense that the 8u2's would be damaged when the processor wasn't, unless it's the size of the chips? The 8u2's are a lot smaller and if something bad was happening in power supply land then their smaller size might have dissipated less heat? And caused them to self-destruct before the processor did?
There are mysteries here folks. Could the the design team of the mega r3 please run some tests using dual supplies or talk about the circuit theory issues that are relevant here please?
The power supply I used was an OKI-78SR 5v dc-dc regulator being powered by a wall wart 12vdc power supply which measured about 16vdc unloaded, and about 15v when plugged into the arduino/shield stack.