I agree that notebooks are not the best in production; better integrating them into your data pipeline. But that takes time, and if users are changes the notebooks still, you end up maintaining two, the pipeline and the notebook. Therefore if using notebooks, the best you can do integrate them with papermill in the pipeline as a step. Plus adding proper data validation and run history etc., around.

Yes, it is most matured for sure, but that doesn't mean it the best solution. If you look, e.g. features and releases dagster or prefect have lately, then in case your starting new, I'd very strongly recommend having a look at them before starting with Airflow.