[exrook]

Why not instead modify the test so that it more accurately measures the capabilities of interest instead?

Such as in your elevator example, make it so that the operation of the elevator is tested with a heavy load approaching the desired safety factor, and evaluating the performance under this heavier than normal load? The idea being that the only way to "cheat" is to design a better elevator that can still meet the requirements.

Of course the reason this isn't done is because of the extra cost of performing a more comprehensive test, but given how much students are paying for education, surely we could make some improvements over the status quo.

I can't help but draw comparisons to SpaceX's starship development process vs that of the more traditional aerospace manufacturers. There isn't any way for the engineers to cheat because their work is tested under the target conditions and either passes[0] or fails[1][2]. The goal being to achieve the most efficient design that still meets requirements.

[0] https://twitter.com/elonmusk/status/1259344535991140352

[1] https://www.youtube.com/watch?v=QlDBjHa0NkU

[2] https://www.youtube.com/watch?v=wFXQ5SRCy74