Dump every captured file + schema change to CSV over a window you pick. One row per change, with a diff URL and blob SHA so an auditor can independently verify the "before / after" content lives in BLOB_STORE.
change_id, kind, captured_at, server_name, target, status, blob_sha, is_ts_only, diff_url, change_summarydiff_url for the full before/after content; blob_sha is the content-addressable pointer into BLOB_STORE for independent verification.