Because I have been unable to reproduce the problem without the hardware, I can only assume the hardware is playing a role here. Unfortunately, the problem only occurs with other hardware connected to my PC and when I access all the hardware from my PC application. For example, the 200,000 overnight test was with all the same hardware and software, but the software was only testing serial communication with the Arduino...
I have previously posted what's connected to the Arduino, two custom shields, one providing 5V - 5A regulation and the other is an LED driver. Likely some combination of the hardware is the problem as you stated, I can reproduce the problem with the simple program I've posted.. As stated previously in this thread, the USB hub could be part of the problem...
I can't help feeling there may be some fuzzy thinking going on if the above represents what you believe to be the entire picture.
The phrase "but the software was only testing serial communication with the Arduino" suggests to me there may have been other factors involved when the failures occurred.
And the phrase "I can reproduce the problem with the simple program I've posted" seems to be a very clear indication that the problem is not within the Arduino. And it seems to make the comment about what's connected to the Arduino irrelevant.
...R