I had occasion to work with a minor investigation into a guarded switch (the red flip-cover type) where letting the cover snap closed would result in the switch changing state. Mainly it was that the sheet metal enclosure was carrying the shock load from letting the cover snap shut into a rotation of the switch body and this would dislodge the switch lever.
---
On another case we had a radar station and the operators kept reporting that the cursor on the screen would randomly wander across the screen. It was very intermittent.
The EEs kept looking at all the wiring, the software guys were stumped. Our company was responsible for the radar system, but another company was responsible for the station (among others) that the radar system mounted to.
Eventually we got some engineers on board the USAF test flights because there was no hint of a cause. It took a while to spot the problem, but it turned out it was the radar operator. The station had a desk-like surface and the cursor control stick, a force/strain gauge system, was mounted underneath with a hole for the stick to poke through. The connector on the side of the control stick box exited the side of the box (pretty sure; it's been 30 years).
If you got to this point, have you spotted the problem?
It turned out the operator would get a bit fatigued, lean back in the station chair, and then put one ankle on the other leg's knee and - dun, dun, DUN! - brace their foot on the connector.
Being a strain gauge control, any deformation of the housing was picked up as an input to the stick. It wasn't a large input; the cursor would just start creeping across the screen. If the operator wasn't watching, and they were fatigued after all, the cursor would appear to them to have jumped to a new spot. They would snap to paying attention and the cursor would cease motion.
This was really a problem to be solved by the installation company who designed the station - but parallel subcontractors get stuck with whatever the prime contractor wants and so I got to design a box to go around the control stick box so that the operator couldn't kick the connector with a boot anymore.
---
For a human factors problem - we had a system mounted to the roof of a vehicle. It was very important that the system was locked down before moving the vehicle. There was a manual latch to do this locking. All very standard. We were to make no changes to the vehicle wiring beyond drawing battery power. To idiot proof this there was a switch controlling a beeper that would sound if the system wasn't in use and the latches weren't locked. Unfortunately this could only sound if the system was turned on.
I suggested this was a bad idea, but management managed to convince the customer that this was OK and that their troops could be trained to operate the latch.
They did not operate the latch and the system fell off because they were moving the vehicle, not operating the system.
I had wanted to use a push-pull cable to cover the 6-8 foot distance, attached to the latch, and operating a flag on the dashboard that would flip down and have "UNLATCHED" on the flag, possibly covering the speedometer. Since the flag would move out of the way when latched, the speedometer would be uncovered whenever the vehicle was ready to be driven. No electricity, no wiring change.
Since the EEs were running the project, they didn't do that.