“Evidence-based” is not a stick to beat other people with. Scientific literature is always subject to review: you cannot assume that because something is written in a paper, it’s proved and decided. Papers are part of the discussion, not the end of it. If you are implying that finding problems with a paper means being unscientific, I think you are misguided.
You cited a paper titled “On the Relation Between Unit Testing and Code Quality”. The paper could not find a strong link between unit tests and code quality. I think there are multiple problems with the paper, from the fact that code quality is not an unequivocal physical measurement like temperature, to the small sample size, making generalization at least dubious. Even then, the result is not surprising: I argue that testing is not about code quality, but about keeping cost of change down.
I would not be surprised to find that eating fruit and vegetables does not reduce the risk of getting struck by lightnings, but that does not mean that we should all stop eating them.
Also, beware of your own selection bias when cherry-picking studies.
