We asked a question that we had been avoiding: if the edge disappeared tomorrow, how long would it take for the results to tell us?
This is a question about statistical power, and it has an answer. Given the size of the effect and the variability around it, you can compute how much data is needed before a dead strategy becomes distinguishable from a live one.
The answer for our numbers is measured in years, not months, and at the pessimistic end it is a long stretch of years.
That is worth sitting with. It means the honest position is not "we will notice when it stops working". We will not, at least not from the returns. Anyone who claims otherwise about a strategy with our shape has not run the calculation.
So the response has to be structural rather than vigilant. Size is set assuming we might be wrong for a long time without knowing it. Capital is held with room. And the things we monitor are the mechanisms rather than the returns, because a broken mechanism shows up immediately and a dead edge does not show up for years.
Nothing about this is comfortable, and it is the single fact that most shapes how we run.