So another industry story from the same effect:
We had an instrument, measuring voltages with some 0.003% accuracy, with some kHz speed. So obviously, it had a precision reference voltage, which was filtered quite a lot. The issue was, that it was picking up any vibration of the PCB, and there were fans in the system (loud ones). So I traced down the issue to the ceramic capacitors filtering the reference voltage. There was a PI filter, CRC, 10 uF, 10Kohm, 10uF (from the top of my head) on the output of the reference voltage, X7R capacitors. That is an 1 Hz filter, going to an opamp for buffering. So when I tapped the capacitors with a pen, they were creating up to a few mV (!) of noise on the output. That is how much noise these can generate.
The caps were replaced with SMD film capacitors. Spare no expense. All the noise was gone, I could hammer those all day long. I guess that was lesson learned.