I have an implementation of an LMM+ model (a shifted Libor Market Model with rebonato volatility function) and am seeing recently that the calibrated parameters are becoming more unstable over time; I have been observing this mostly since the war in Ukraine started. In terms of the implied volatility surface, we have seen that the war has led to an inversion of the volatility term structure; i.e. whereas in the past we usually saw higher vols for further maturities, the opposite is now present. Basically, it's becoming more expensive to hedge against IR rises in the short-term, which makes sense given the behavior of central banks. But I am wondering how this could explain the decreased parameter stability over time. Anyone any idea how I could go about to investigate this? Many thanks!