two message legs tell the author the reader refuses and skips their record when the admission reader counts it and the surprise store has no reader at all

two message legs tell the author the reader refuses and skips their record when the admission reader counts it and the surprise store has no reader at all

Found by the round-3 ruthless review. FIX-FIRST. The fourth instance on this branch of one class: a message asserting a mechanism that is not the case.

checkRecordUnknownFields (schema.go:786-788) emits, as a BLOCKER, that a key outside the schema "makes this a record the reader refuses and skips — invisible to every admission surface while it still sits in the store". Probed: ReadReadingOutstanding returns Unadmitted=[] — the record is not skipped and not invisible, it is ACTIVELY COUNTED. For srp the message is worse: no surprise reader exists anywhere in the tree, so "invisible to every surprise surface" names an empty surface set and a refusal nobody performs. The author is sent to look for a record that is being read.

checkRecordDuplicateKeys (schema.go:1201-1203) carries the same claim for EVERY store: "the record reader refuses a duplicated key, so the file is skipped by every admission surface". True for iss (capture/parse.go:122); false for adm, srp, and — pre-existing — rdi, rdg, dsp.

The sting: 6226f7d2 fixed exactly this defect in checkRecordRequiredFields one function away, and did not check the two legs beside it. The flag it added, readerFailsClosed, is correct for all four stores on the UNKNOWN-KEY leg, which is the leg both reviewers checked: adr declares no knownFields and returns early, so the claim is never rendered there.

It is NOT correct for adr on the DUPLICATE-KEY leg, and the reviewers carried their unknown-key verdict silently across to it. record.readRecordHead reads with frontmatter.Fields, the lenient scanner, and no ADR reader anywhere refuses a duplicated key: an ADR carrying status twice renders, with the first value and a nil error. The duplicate-key leg is declared separately, as readerRefusesDuplicateKey (iss-2608301656200729).

Remedy, one condition, both sites: gate the reader clause on r.store.readerFailsClosed exactly as checkRecordRequiredFields:745 now does, keeping the schema-closed half unconditional. adr is unaffected on the closed-schema leg (nil knownFields, early return) and must not keep the claim on the duplicate-key leg; iss keeps both, which is true of it. Then widen the flag's godoc: it covers required properties and the closed schema — one declaration, two legs, rather than one leg declaring and one assuming — with the duplicate key on its own declaration beside it.