PR #116 at f2a981d: test change covered; MONMM wrong answer blocks

Author: kvlonge

coverage/syntax/audits/launch-blockers-pr116-f2a981da.toml

@@ -0,0 +1,16 @@
+# D260 carry-over check for PR #116: fix/d250-launch-blockers f2a981da against the approved be38679f (record launch-blockers-pr116-be38679f.toml).
+
+[review]
+schema = "sqlquack.recorder-change-review.v1"
+decision = "D260"
+branch = "fix/d250-launch-blockers"
+commit = "f2a981da02d2d267b867c643f58a05b8d618b6be"
+previous_head = "be38679f"
+reviewed_at = "2026-11-06T11:58:14Z"
+verdict = "changes_requested"
+checked = "926536fe has the same tree as be38679f (f2863395). f2a981da changes one line of test_coverage_batch.rs: the 'another base refuses' row uses POWER(3, 3) instead of POWER(2, 3). The expected error 'other than the constant 10' is still contained in the new refusal text ('other than the constant 10 or 2'), so the row still checks the refusal. The test change is covered by the be38679f approval. My round 1 ran only the format test, so it did not catch the POWER(2, 3) row failing."
+
+[[review.finding]]
+kind = "wrong_answer"
+blocking = true
+detail = "A's receipts show Snowflake writes TO_CHAR(date, 'MONMM') as the full month name, but #116 converts it as MON then MM ('Sep09'). Main refuses MON, so #116 turns a refusal into a wrong answer. My round 1 listed MONMM as unrecorded only. Smallest fix: take the follow-up's refusal of any field directly after MON into #116, or land that refusal in the same slot. Also check MMMM followed directly by a field (MMMMDD) against a receipt, or refuse it, since Snowflake does not read MON adjacency the way #116 assumed."