I was troubleshooting a digital power meter ("Watt-A-Duino", based on Arduino and dsPIC) that I made for senior design. The calibration drifted badly with the 5V supply. I only needed to operate it on PoE for the demo, so I didn't bother looking into it. But now that I need to actually use it, I needed to figure out why the voltage channel was getting such a poor PSRR. The AFE opamps showed no change as the 5V supply varied a little. I was puzzled since the voltage channel used the dsPIC's internal ADC, which ran off the 3.3V supply that did not drift much. Turns out that in doing level translation between the dsPIC and Arduino, I just put in series resistors. That did limit the current to a safe value, but the current flowing through the protection diodes caused charge injection that was messing up the ADC. I added some clamping diodes and now the PSRR is as good as I expected.
In another part of the project, I had a via come apart (probably blew like a fuse) and start arcing. Things get pretty exciting when it's on a 270V DC bus with 440uF of capacitance!
http://arduino.cc/forum/index.php/topic,80200.0.html