QM was literally one of those "huh, that's funny" type of moments piled on top of others, like starting from the classically troublesome but sorta reasonable assumption that a box with radiation bouncing around inside could have a smooth distribution of wavelengths present within, instead of the distribution you would only expect if you started by assuming you can have 1 foo of energy or 2 foos but not 1.389 foos.
Very rarely is there ever a situation where some lone enlightened genius goes "ah hah, this thing is actually acting this way because of that other thing!" and sets off a whole new round of old bearded dudes agreeing and congratulating them.
You tend to see somebody going "huh, wasn't this supposed to output a signal like this when we feed this input? why is it doing this whole other thing with this one range of inputs? let's check with that other one over there and see if it's time to get a new thingamajig" before you end up having to start throwing guesses at the wall, and if you're lucky you can construct one well enough that it ends up accurately describing other observations and predicting entirely new ones.
Science is a way to avoid being wrong about stuff, yes it is true that a single experiment can invalidate a whole stack of previously valid assumptions, but we've been picking off those sorts of low hanging fruit for a long time now and searching for new results isn't always as important as confirming old ones. Indeed if you aren't trying to make sure others can perform the same experiments and reproduce your results then you might be doing something that looks like science from the outside without noticing all the different ways you may have fooled yourself.
Without being able to check for that, you run the risk of passing on those same issues so they can fool others, as exemplified by the
replication crisis snowballing up into such a clusterfuck of research building on prior results under the assumption that they could be reproduced and thus it was a waste of time to do so.