Simpler to read: Running KS tests does not "guarantee" statistical validity; the quality of the reference data remains crucial.
The simplest is that poor quality reference data distort any test conclusion, no matter how sophisticated.
Without rigorous validation of inputs, any result of statistical similarity is deceptive.
For example, tests performed with US market data on an Asian portfolio like the MAS would produce meaningless conclusions.
The parsimony dictates solving the data quality issue first.