Let's avoid rushing to conclusions about the concretization of statistical similarity via Kolmogorov-Smirnov (KS) tests when empirical data are intrinsically biased.
The exit condition comes first: if reference distributions are built on abnormally calm market periods, this masks real downside risks.
A statistical similarity does not protect us against a localized tail risk if the global benchmark does not account for sector-specificities or provincial regulations.
For example, a Canadian natural resources portfolio compared to a global benchmark that omits systemic shocks specific to Canada does not reflect the reality of the mandate.