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