Your interest rate momentum strategy was highly profitable for the last decade but had a massive drawdown recently. Looking at the time series of the 10-year Treasury yield, what statistical properties likely broke down? How would you redesign the model to detect such regime shifts in real-time?