One sentence: performing Kolmogorov-Smirnov tests only realizes statistical similarity if the underlying hypotheses are solid.
Cut the noise: the parity of distributions depends entirely on the relevance of input data.
If the benchmark is biased, such as using post-crisis historical data for pre-crisis forecasts, the KS test validates an incorrect comparison.
It is the simplest method, but it does not correct defective data.
For example, validating the similarity between venture capital portfolio returns and Gaussian benchmarks is a false security, ignoring real extreme risks.