readerFailsClosed is false for the adr store on the duplicate key leg and the record that resolved it claims all four stores were verified

readerFailsClosed is false for the adr store on the duplicate key leg and the record that resolved it claims all four stores were verified

Found by the round-5 ruthless review, and it lands on a record this branch already resolved.

readerFailsClosed is set on the adr store and gates all three legs on the godoc's assertion that "The three malformations differ; what the reader does about them does not." That is false for adr on the DUPLICATE-KEY leg. record.readRecordHead reads with frontmatter.Fields, the lenient scanner, and no ADR reader anywhere refuses a duplicated key.

Probed: an ADR carrying status: accepted then status: draft renders -- describeADR returns the record with Status:accepted and a nil error -- while the gate says "the record reader refuses a duplicated key, so the file is skipped by every ADR surface".

The part worth keeping: iss-2608301519254418, resolved on this branch in round 4, asserts the flag is "correct for all four stores (verified independently by both reviewers)". Both reviewers verified the UNKNOWN-KEY leg, where adr has nil knownFields and returns early, and carried that silently across to the duplicate-key leg, which does fire. A verification claim in a resolved record is exactly the thing a later reader trusts instead of re-deriving, so the record needs correcting alongside the code.

TestClosedSchemaAndDuplicateKeyClaimNoReaderWhereTheStoreHasNone covers adm, srp and iss, and not adr.

Remedy: split the declaration so adr does not stand behind the duplicate-key clause, and extend that test to the ADR store.