The shortest explanation: interpreting a statistical test is not an proof in itself; it is conditional on the quality of the data.
A Kolmogorov-Smirnov test showing similarity does not guarantee performance robustness if the inputs are faulty.
If the data are biased, for example, hedge fund returns from Hong Kong without adjustment for capital controls in Singapore, the similarity is misleading.
The test does not validate the interpretation, only a mathematical measure; implications require parsimony.