LibriVox reader-credit sequence audit Source snapshot: 31 October 2026 UTC. Ten selected catalogue projects, five paired novels, 321 listed files. A credit ID is a catalogue identifier, not independently verified unique person or audible voice. Selection before section-credit outcomes Six novels and fixed Gutenberg eBook IDs were declared: Jane Eyre 1260, Frankenstein 84, Moby Dick 2701, Alice 11, Great Expectations 1400, Secret Garden 113. For each work, select the two lowest numeric project IDs among matching English complete catalogue projects with that documented source-eBook identity, excluding explicitly abridged, dramatic, selection and anthology projects. Prefix-discovery result sets were all below the requested 50-record limit. Moby Dick had no second matching 2701-source project; it was retained as an unmatched discovery, not replaced by the later 15-source project. The main sample is the other five pairs. Numeric ID order is not release chronology. The minimum was eight fully section-resolved projects across three works, fixed before metrics; the observed source gate passed ten across five. The frozen private candidate ledger is preserved. Independent QA detected an error in explanatory strings for ineligible candidates: a producer comprehension listed all four possible failure reasons. An append-only correction recomputed the actual failed criteria for all 46 candidates. Eligibility and selected-project flags were unchanged; no work, project, source group or outcome was replaced. Factual public inputs project-facts.json retains the ten selected IDs, original catalogue titles, source links, catalogue dates, header-credit interpretation and declared file counts. section-credits.csv retains all 321 section IDs, original numeric section numbers, ordinal positions, exact reader-ID sets and display names. The public Read by label LibriVox Volunteers is a collective label, not a single individual or a complete numbered roster. Role credits are excluded unless present in section rows. Sources and dates are in sources.json. API and catalogue HTML are two representations from one provider, not independent recordings. Raw private response hashes identify original captured bytes; they are not audio hashes or public-recipe dependencies. Calculations Sort each project by unique integer section_number, retaining zero. A known-set adjacency requires both adjacent credit sets to be present. A handoff is a difference between those exact sets. Report changed/eligible boundaries with rational half-up rounding to one decimal percentage. A singleton adjacency separately requires one ID on each side. All captured sections are resolved singletons, so both measures coincide; missing/multi-ID states are supported explicitly and never imputed. A run is a consecutive stretch of the same singleton ID. Retain start/end ordinal positions, original section numbers and length in files. Nonadjacent returns mean an ID appears in more than one separated run, not a new individual. Per-ID counts preserve every occurrence separately from runs. For fully singleton-resolved projects, a hypothetical regrouping of the identical ID multiset into one block per distinct ID attains the minimum K−1 handoffs. This mathematical alternative is not an observed recording, a proposed narrative edit or evidence about quality. Portable reproduction Copy reproduce.py, project-facts.json and section-credits.csv together into a fresh INPUT directory. Run Python 3.10 or later, standard library only: python -I -S reproduce.py --input-dir INPUT --out OUTPUT No network, credentials, packages or private workspace paths are needed. Compare all six generated files: project-metrics.csv, adjacencies.csv, singleton-runs.csv, reader-credit-counts.csv, sequence-chart-points.csv and summary.json. SHA256SUMS.txt pins the selected public inputs, code, outputs and documentation; it does not include itself. UTF-8 and LF are explicit. A later live catalogue can change without altering this frozen reference. Limits and prior art No audio, novels, covers, performance ratings, listener outcomes or duration-weighting used. A file need not be one chapter; shared source links do not verify historical text bytes or spoken content. Purposeful early-ID selection and possible shared/reused material prevent representative or independent-performance claims. Existing solo/group filters, section credits and shared-source grouping are prior art; original contribution is the inspectable ordered boundary/run reference and its contributor-count rank reversal. The provider’s obsolete pre-1923 policy explanation is not adopted as current legal guidance. No worldwide rights adjudication is made.