Well, this doesn't have to do with a FIRST robot, but it was certainly a strange issue... and the robot it was with was indirectly responsible for me becoming a mentor
.
Back in college, I took an autonomous robotics class. Basically, create a robot from Legos, motors, sensors, and a provided logic board to play a game involving searching for brightly colored plastic eggs and returning them to a nest. Got quite a ways into development of our robot, and were testing it out in the lab, along with several other groups.
Anyways, our robot was having some sort of programming issues... it was seemingly randomly going off-script and acting in ways that weren't actually possible according to static analysis of the code. We started adding in some print statements (there was a small display on the logic board allowing maybe a dozen characters to be displayed at a time). As we added more and more print statements to see what it was thinking, the behavior got worse and worse - literally the only changes were some print statements, yet it was going off-script faster and faster as we tested.
After a few days, we found out this was a bug with the display buffer on the logic board that the professor never bothered to tell anyone about - something about the way it worked behind the scenes caused the memory on the board to get corrupted if you tried to display an odd-length message. So with each print statement we added, there was a 50% chance it would corrupt the memory (as the message could have either an odd or even number of characters). We went through and "even-padded" every print statement (including the one that originally caused the problem) and the robot worked perfectly without any other modifications.