Actually, pulling out the USB cable was the main method I used to fix it before I discovered I could restart the computer.
I believe this implies something isn't right in the usbser.sys driver since XP. Another link I posted implied this problem went away when using a XP machine.
Less noise would be better, but I can't guarantee the noise levels in the background when I deliver this instrument. While it could be true that the many cables I tried were inadequate, I don't believe this explains why the problem goes away when I use the USB 2.0 hub instead of the USB 3.0 hub (unless the two USB 3.0 A->B cables I have are also bad). The arduino cable is already very short at 18".