Hi @valiro21 ,

You are correct that in your case APPEND ONCE flow won't work as you have MV as target. 

The scenario that you provided is quite specific. Most of the times, if the MV logic changes, the new logic will require to look back to the source data and recompute the aggregates. The engine is not aware of the fact that avg and median are close values for your dataset. It will need to recompute anyways to keep the promise of the MV being "always correct". 

An option would be to include both metrics (avg and median) in the initial MV to avoid changing the MV definition.

Another option: create a new MV with new logic and union it together with the modified values from the old MV.

Downstream MV will indeed get updates of the upstream MV and will potentially need to recompute old values (this can happen incrementally). You can perform tests and check how it behaves.

Best regards,