Scientists are human and fallible.

"Professional Science" is just as vulnerable to "eh, I know what I'm doing", bias, politics, funding, feuds, ignoring details-that-dont-fit and shortcuts, as the rest of the human experience.

That's why we see "breakthrough discoveries" falling apart to scrutiny on a regular basis and new facts/theories are only gradually accepted into the "body of accepted knowledge" after lots of peer reviewing, reproduction, general chewing-it-over and when the old "that can't be true" generation has retired/died.

On the other hand, quick and dirty gut-check experiments and goofing around with a new idea are a valuable way to easily check for falsification and narrow down what actual, rigorous tests might have to look like. They're also a major source of lab accidents.

In the context of the Manhattan Project, the demon core is a perfect example of this.