Assume a person who believes that testable ideas are the most interesting ones. Or (weaker form) that stuff that can be laid down in a very neat and logical way or is testable has a strong superiority.
He will devote all his time etc. to this kind of questions, thereby creating a whole body of knowledge that has been selected not on the basis of importance, utility, reality & truth, but on the basis of what meets the technical criteria.
This may create a very distorted world view. and a off the mark body of knowledge.
Even more extreme oddities will emerge in the dynamics of knowledge and learning. When there is more literature and more people talking about a minor question that happened to fit the "science" criteria. Now other people will develop their life, perception and research in this direction, deviating even more form the more relevant issues that do not fit into technical handling.
These dynamics may create a very biased world view. As examplified elsewhere, small innitial biases may contribute to great deviation in the end.
A great example can be seen with happiness research. Put aside what is "correct" for happiness, it seems plausible that even just because some happiness measures are more "neat" and handy, they become "the currency of trade" in the research community.
AFTER they have become "the currency" of the research, they have several advantages including "framing" (i.e. they are written on the wall of the mind all the time as happiness strengthning their imagined relavance). The other advantage of is that even if you disagree with the idea, you will use it just because there is so much research based on that.
On balance, we have many reasons to value technically collected research. We should as well remember that our goal is not being technically correct. We want certain things from life, and should remember that technical science and logic can handle only part of reality.
Reality is there as it is. It does not care what we want it to be. Sometimes, reality even fights back to those trying to ignore it.