I did some further reading and got the same conclusion. APPLY CHANGES might to the trick. However, I don't like the limitations. From Bronze to Silver I might need .foreachBatch to implement the JSON-logic and the attribute names (__start_at / __end_at) seem to be static - this can make migrations of BI-layers (gold) hard

Anyway, I think you answered my question by confirmation. Since I'm doing this for learning & fun, I'd like to develop my own "Databricks-native Datavault 2.0"-Loader with respect of PII-handling as thought in the training 😉