Ha, no I hadn't seen this thanks for flagging it -- I can't provide a deep debrief right now but fwiw I understand where at least some of his objections come from. I'll try to give it a more earnest read later. My claims are based on intended use of the Airflow architecture; for example, things like hacking the scheduler (which re-runs your DAG code with every pass) does give you some dynamicism but that doesn't make it a feature (or a good idea)