I was working for an avionics maintenance facility where a variety of military avionics systems were tested, repaired, and overhauled. I was part of the team that provided engineering support for the repair floor. My job was to troubleshoot test software and hardware, provide tooling, and sometimes solve problems with incurable chemical concoctions. I had to solve the problems that no one else could (or wanted to).
On one assignment, I had to handle an unusual problem with a temperature-controlled oscillator that was being used as a radar receiver. The automated test program that was testing the receiver reported sideband amplitudes well above the maximum test limit. The technologists had no idea why.
They removed the faulty oscillator and replaced it with another, but the problem persisted. They removed the oscillator from the receiver and performed a bench test using a power supply and a spectrum analyzer. The bench test produced perfect results on both test oscillators, with all sideband values within limits. The oscillator was then reinstalled into the receiver, and the problem returned.
They experimented on the bench with varying the load and heat cycling the oscillator, to no avail. When the problem came to me, I performed their testing methodology, and it appeared to all of us that the only thing left to examine was the radar receiver power supply. I asked the technologists if they had checked the quality of the power supply. Obviously annoyed that I assumed they hadn't checked the power supply, they assured me that it was nice and flat, like all DC power supplies should be. They went further to say that the ripple value was in tolerance, as well. I felt a bit apprehensive about posing further power-supply inquires.
We then explored the possibility that there might be a grounding issue, since the frequency of the oscillator was quite high. We moved the power supply from the receiver and tried a variety of grounding techniques, with no success. After some time working this job, you get an intuitive sense of where to look.