I get that. But my point is: are we really sure that this is the problem?

One of the bases of our scientific method is repeatability of experiment. But at some point, when we can produce a lot of experiments, comes a problem: we can run out of people with time and resources that allow to repeat it. And one of the ways to mitigate it is to strengthen the requirements on the data gathering. So when you do find something weird, you can analyze how the parameters differ from other similar runs and if someone else is able to repeat it, you might have easier time finding which variable makes it so. Without consistent "we measured X after setting that to Y" it's hard to repeat the experiment or even recognize if you really are observing something new.

Take a look at that error a few months ago that resulted in us thinking that a new superconductor that can work in +- room conditions was found. If we didn't have precise description of what they did and what they measured, we could be still trying to reproduce their observations