In the 1970s there was a major shift in how aerospace engineers treated the fatigue of metal parts.
Before the shift, the engineers assumed the part was manufactured to print with the specified surface finish and then the calculated or measured stresses were "looked up" on a table and the read-across was the number of cycles the part was expected to survive.
That changed after some very high-profile structural failures of wing-knuckles. Post-mortem evaluation revealed that the fracture started at a scratch or gouge left by a repair tool that had slipped or fallen.
After those failures, the preferred mode of analyzing parts was to assume a crack already existed and to use a crack-growth-rate model to see if the crack would significantly reduce the load carrying capacity of the part during the air-frame's expected lifetime.
Life was good for a while but then more cracked parts were found and the cracks were growing faster than anticipated.
After putting their heads together and coming up with a plan, the design engineers called in the test pilots (who also had engineering degrees) and explained that they thought that the problem was caused by high-frequency, mechanical vibration. The thinking was, the design engineers explained, that the high frequency resulted in many, many cycles being added to the part very quickly. The design engineers said that conventional instrumentation had analog filters to filter out high-frequency content due to limitations on being able to record all of the data.
So the design engineers had designed special instrumentation to capture the high-frequency content. They had overcome the data limitations by installing a switch which the pilots were told to turn on just before they performed a standard limit maneuvers and turned off immediately after completing the maneuver.
The story about the switches was total bull-shit. The devices were recording the entire time.
What the design engineers learned is that improvements in flight-suits and "grunting" meant that standard "design limit maneuvers" were no longer at the limits of human performance. The pilots (thinking that nobody was watching) exceeded "limit maneuvers" on the way to and from the desert where such maneuvers were officially performed. They flipped the switches and generated the expected g-levels and then flipped them off. The interesting events were on the way to-and-from the desert.
Some personalities are more likely to be non-compliant than others. Some of us are non-compliant in creative ways.
Incidentally, the rototiller started on the first pull. The only unexpected event is that a bee (yellowjacket? bumblebee?) stung me on the back of my calf. Of course it was the leg that had been broken.