Tbh the article seems to just be like "you can't solve idempotency with one idempotency-key header" and well like no shit.

If the atomic action is idempotent, you don't need a layer for repeating yourself. You hit the nail on the head. So much idempotency efforts are made because they never made the actions idempotent in the first place.

[dead]