Five validation lanes. One release boundary.

Across 32 products, 26 were contract-complete and the same six reported warnings were reproduced with zero false positives and zero false negatives. grid-source-registry and grid-edge-conditions lack observationDate; constraint-consequence-ledger, rooftop-forecast-record and der-adoption-ledger each lack currentVersion and observationDate; export-flexibility-dividend lacks currentVersion.

Method: Applied the same presence rule used by the trust register: sourceLabel, currentVersion, coverage, cadence, dataConfidence, knownLimit and at least one of observationPeriodEnd, latestObservationDate or latestReleaseDate must be non-empty.

Evidence: Live data-status register (2026-08-02); Live data-trust register (2026-08-02); Trust completeness implementation (2026-08-02)

Boundary: This check assesses shared publication metadata only. A schemaVersion or methodVersion on a route is not treated as a released data edition, and metadata completeness does not prove source accuracy.

All 19 declared exports returned HTTP 200; JSON and schema routes returned application/json and CSV routes returned text/csv.

Method: Issued read-only HTTP GET requests to every JSON, CSV, schema and health export declared for the six products and compared response content types with the declared format.

Evidence: Product export declarations (2026-08-02); Grid source registry export (2026-08-02); Export flexibility export (2026-08-02)

Boundary: HTTP 200 proves route reachability only. Three JSON routes deliberately return status=unavailable, so this result is not evidence that their source datasets have been released.

Twenty-seven of 32 products expose no public non-contract release receipt. Four products have pass-only receipt sets, solar-capacity-pipeline has one warning receipt, and all six target products have zero receipts. Route-level source receipts, where present, have not been promoted into the shared trust register.

Method: Counted public data-trust releaseReceipts after excluding the contract-completeness receipt across all 32 products, then isolated the six warning products.

Evidence: Live data-trust release receipts (2026-08-02)

Boundary: Absence from the shared trust register does not prove that no internal validation occurred; it means the public release cannot currently recover an inspectable verification receipt from that register.

grid-source-registry publishes sourceVintage 2026-08-02T00:00:00.000Z but no shared observation date. grid-edge-conditions publishes editionVersion grid-edge-2026-07, periodEnd 2026-07-31 and releaseStatus published, while its shared record has no observation or release date.

Method: Compared the edition and vintage fields in each product's public JSON with currentVersion and observation-date fields in data-status and data-trust.

Evidence: Grid source registry (2026-08-02); Grid edge conditions (2026-08-02); Shared status projection (2026-08-02)

Boundary: The route fields are inspectable evidence of a vintage or edition, but the audit does not choose which shared date column is semantically correct; that mapping must be explicit in the parent-owned release pipeline.

The local public schema is export-flexibility-dividend-public-v2. Every result requires publication.state=interactive_model_only, canonicalBenchmarkPublished=false, durableReleaseStatus=provisional, computationalVerification.status=passed with checkedOn=2026-08-02, and empiricalValidation.status=pending with no cohort edition and an explicit limitation. The former ambiguity between calculated output and canonical publication is resolved.

Method: Inspected the local v2 result contract, JSON Schema, CSV projection and interface copy, then ran the focused model and public-contract tests to verify that every calculated result carries the same mandatory publication boundary.

Evidence: Local v2 result contract (2026-08-02); Required JSON Schema boundary (2026-08-02); Publication boundary tests (2026-08-02)

Boundary: This pass verifies truthful publication semantics and computational checks only. It does not fill the shared currentVersion field, create a public non-contract release receipt, publish a canonical benchmark or establish empirical accuracy and transferability.

constraint-consequence-ledger, rooftop-forecast-record and der-adoption-ledger each return status=unavailable, a specific source-integration reason and an empty sourceReceipts array; none substitutes example or zero-filled observations.

Method: Inspected status, reason, editionVersion, sourceUpdatedAt and sourceReceipts on the three source-dependent public JSON responses.

Evidence: Constraint ledger response (2026-08-02); Rooftop forecast response (2026-08-02); DER adoption response (2026-08-02)

Boundary: The method and response contracts are publishable, but the three data products remain recovering and unresearched until receipt-pinned source editions are integrated.

All nine profiles were audited: one Ready, four Caution and four Hold, with no profile promoted by the validation layer.

Method: Ran the executable decision-readiness diagnostic over every profile, product and release receipt returned by the live data-status endpoint, retaining assessDecisionReadiness as the state authority.

Evidence: Live data-status contract (2026-08-02T03:17:50.952Z); Executable diagnostic (2026-08-02)

Boundary: The fixed result describes the named endpoint snapshot. A later source-health transition can change readiness and requires a new dated run.

A required recovering input produces Hold. The four live Holds each resolve to a required operations blocker, not an optional-context or bookkeeping issue.

Method: Executed fixture tests with a required recovering product and separately inspected every live required input marked Hold.

Evidence: Decision readiness sprint tests (2026-08-02); Existing readiness method (2026-08-02)

Boundary: The diagnostic explains the existing method result; it does not alter the profile registry or override product health.

Optional degradation produced Caution, never Hold. Six optional-context blocker instances were classified with their operations or confidence root cause retained.

Method: Executed fixture tests with all required inputs Ready and one optional recovering input, then compared the result with live Caution profiles.

Evidence: Decision readiness sprint tests (2026-08-02); Live data-status contract (2026-08-02T03:17:50.952Z)

Boundary: An optional input can explain caution only while every required input remains usable under the existing readiness method.

Thirty of 34 profile-input instances matched an exact receipt. Product-Market Evidence, SWIS DER Adoption, Export Constraint Regime Register and Australia Grid Edge Conditions had a missing or mismatched direct receipt trace.

Method: Matched each profile input's productKey and currentVersion against the release ledger, preserving missing and version-mismatch outcomes as metadata findings.

Evidence: Live product and release ledger (2026-08-02T03:17:50.952Z)

Boundary: A missing direct receipt is a registration gap unless separate evidence proves the underlying source absent or invalid; it is not automatically an evidence failure.

Thirty-three of 34 profile-input instances satisfy their registered date basis. Australia Grid Edge Conditions is the only unresolved instance because its source_check basis has neither sourceCheckedAt nor latestRetrievedAt.

Method: Evaluated one accepted date alternative per product freshnessBasis: sourceCheckedAt or latestRetrievedAt for source_check, sourceReleaseDate or latestReleaseDate for source_release, observationPeriodEnd or latestObservationDate for observation_period, and lastMaterialReleaseAt, latestReleaseDate or latestRetrievedAt for editorial_review.

Evidence: Live product status ledger (2026-08-02T03:17:50.952Z); Freshness basis product contract (2026-08-02); Basis-aware regression tests (2026-08-02)

Boundary: Dates belonging to another freshness basis are non-applicable and are neither required nor reported missing. An unknown basis remains not tested rather than borrowing alternatives.

All supplied required dependency rows resolved to registered products in the snapshot, but the public data-status contract does not expose the raw dependency rows used by the server-side health fold.

Method: Joined the eight migration-declared dependency edges relevant to current profile products to the live product status snapshot and inspected required dependency state.

Evidence: Data product dependency migration (2026-08-02); Temporal value dependency migration (2026-08-02); Live product status (2026-08-02T03:17:50.952Z)

Boundary: Dependency completeness was checked against repository-declared rows, not a read of the owner-only production dependency table.

The rerun reports five bookkeeping findings: four unresolved direct receipt traces and one basis-aware source-check date gap. Six universal-field false positives were removed without changing readiness.

Method: Executed fixtures separating an absent release receipt from evidence loss and proving that only the alternatives declared by freshnessBasis participate in date completeness.

Evidence: Decision readiness sprint tests (2026-08-02); Executable diagnostic (2026-08-02)

Boundary: This separation does not excuse missing receipts. It prevents an unknown registration state from being overstated as proof that the source evidence failed.

All five current source families have an explicit source-specific time, interval and unit basis; no cross-provider averaging or implicit unit conversion is introduced.

Method: Inspected the five canonical ingest adapters and recorded each legacy dataset, canonical dataset, timezone basis, interval basis and source unit basis in a provider-neutral source-family register.

Evidence: Source-family register (2026-08-02); Canonical ingest adapters (2026-08-02)

Boundary: The register describes the repository implementation and does not independently certify publisher completeness or licensing terms.

Five synthetic source-family fixtures produced the declared statuses and classifications deterministically: one pass, one warning and three failures.

Method: Executed the fixture suite with node:test and compared repeated receipts for exact deep equality across pass, warning, numeric failure, timezone mismatch and missing-versus-zero cases.

Evidence: Comparator implementation (2026-08-02); Fixture tests (2026-08-02)

Boundary: Synthetic fixtures validate code paths only and contribute zero days to the production comparison gate.

Timezone and interval incompatibilities fail before value comparison, incomplete coverage remains unavailable, and a missing aggregate is not treated as a measured zero.

Method: Exercised explicit compatibility checks for identity, source unit, timezone, UTC day window, interval duration, expected interval count, minimum coverage and value state before numeric tolerance evaluation.

Evidence: Synthetic edge-case fixtures (2026-08-02); Validation coverage (2026-08-02)

Boundary: Per-metric tolerances must be supplied explicitly; fixture tolerances are test inputs and are not production calibration values.

Fixture receipts are excluded from rollout counts; even a passed comparison cannot alter serving mode, and a completed 30-day production gate would still require manual review.

Method: Asserted that every fixture begins with legacy-primary and canonical-shadow roles, leaves legacy-primary unchanged and disables automatic promotion regardless of fixture outcome.

Evidence: Shadow-record implementation (2026-08-02); Serving-boundary tests (2026-08-02)

Boundary: The validation harness does not write rollout state, production receipts or database rows.

The endpoint reports five current sources, zero compared metrics, zero compared days, zero passing days, zero promoted sources and legacy-primary serving for every source.

Method: Queried the public grid-source health endpoint read-only and inspected reconciliation and rollout fields for all five source families.

Evidence: Public grid source health receipt (2026-08-02T03:08:58.709Z)

Boundary: No production comparison history exists, so the 30-day reconciliation gate cannot be evaluated or claimed.

The local Value Under Constraint default returned temporal-value-wa-v1, 11,822.58 kWh annual generation and AUD 962.09726143799 annual value. The corrected local Export Flexibility Dividend returned public schema v2, calculation method export-flexibility-dividend-wa-v1, three interactive scenarios and AUD 127.03/year modelled flexibility headroom.

Method: Executed both repository calculation cores with fixed local inputs, method versions and source editions after the schema-v2 contract correction.

Evidence: Value Under Constraint calculation core (2026-08-02); Export Flexibility Dividend schema-v2 calculation core (2026-08-02)

Boundary: Deployment status was not assessed after the schema-v2 correction. These are local computational reproductions, not a canonical benchmark or empirical site-performance test.

Local contract export-flexibility-dividend-public-v2 retained state=interactive_model_only, canonicalBenchmarkPublished=false, durableReleaseStatus=provisional, computational verification passed on 2026-08-02 and empirical validation pending; all six unsupported promotions failed validation.

Method: Validated the required publication object on the local result and mutation-tested six unsupported promotions: schema v1, canonical state, canonical benchmark published, durable release published, changed computational check date and empirical validation passed.

Evidence: Schema-v2 publication invariant (2026-08-02); Publication-boundary mutation tests (2026-08-02)

Boundary: A passing computational model does not publish a canonical benchmark, complete a durable release or establish empirical accuracy.

The Value Under Constraint default reproduced stable SHA-256 1a6c085fcea51327d42ed2d6037eaacd967fc33057f54035d20d42eea8b51ff6; repeated Export Flexibility Dividend calculations with a fixed source receipt and generatedAt value were identical.

Method: Ran each fixed-input calculation twice, canonicalized object keys, excluded only the declared volatile generatedAt timestamp and compared SHA-256 hashes and deep results.

Evidence: Reusable deterministic hash check (2026-08-02); Model validation tests (2026-08-02)

Boundary: The result is deterministic only while inputs, calculation methods, public schema and source editions remain fixed; generatedAt is presentation metadata and is excluded from the canonical hash.

Three retained lifetime golden cases and 90 year-model combinations passed. The local no-storage default had annual balance residual 5.820766091346741e-11 kWh and maximum hourly residual 0 kWh against a 0.000001 kWh tolerance.

Method: Recomputed every hourly and annual balance as generation minus direct self-consumption, storage charging, export and curtailment. For no-storage cases, also enforced generation = self-consumption + export + curtailment.

Evidence: Invariant implementation (2026-08-02); Golden and matrix execution (2026-08-02)

Boundary: When storage is present, charging is a separate generation destination; collapsing it into direct self-consumption would misstate the engine's published fields.

All 90 deterministic year cases passed; maximum hourly export above a declared fixed cap was 0 kWh. The unconstrained case was treated as a counterfactual with no cap-compliance claim.

Method: Exercised zero, 1.5 kW, 3 kW, 5 kW and unconstrained envelopes across three tariffs, three load shapes and zero or 10 kWh storage, checking every hourly export value against the selected fixed cap.

Evidence: Export-cap matrix test (2026-08-02); Published export-envelope register (2026-08-02)

Boundary: A modelled fixed cap is not a connection approval and does not represent a connection-specific dynamic operating envelope.

The default 8,760 rows summed to 11,822.580000000113 kWh generation and 5,999.999999999818 kWh load; all interval tariff rates matched and rebuilt AUD 962.09726143799 annual value.

Method: Verified 8,760 calendar-hour rows per case, reconciled hourly generation and load sums to annual declarations, resolved each import and export rate from its tariff period, and rebuilt avoided-import, storage and export value from interval flows.

Evidence: Perth solar-shape receipt (2026-08-02); Alignment and tariff invariant tests (2026-08-02)

Boundary: The NASA POWER profile supplies a representative Perth solar shape and the load profiles are synthetic normalized shapes; neither is a measured customer interval trace.

All three 25-year golden cases passed at 0.000001 tolerance. The local default recomputed to AUD 13,439.655099719674 present value, AUD 5,439.655099719674 net present value and discounted payback in year 12.

Method: Recomputed each annual discounted cashflow, cumulative discounted balance, total capital, present value, net present value and first non-negative payback year from the declared settings.

Evidence: Lifetime invariant implementation (2026-08-02); Retained golden cases (2026-08-02)

Boundary: The arithmetic follows the declared model convention: year-one cashflow is discounted once, while tariff escalation, operating cost, finance, tax and storage replacement capital are not modelled.

All three comparison cases ordered fallback <= published ceiling <= unconstrained for annual value and export, and fallback >= published ceiling >= unconstrained for curtailment. A fourth high-capital case confirmed negative net present value remains valid while negative physical flow fails; local interactive comparison arithmetic reconciled within the 0.02 rounded-output tolerance.

Method: Ran three deterministic household and storage cases on one comparison-basis key and checked that broader export access could not reduce export or annual value, increase curtailment, or create negative comparison deltas.

Evidence: Scenario invariant tests (2026-08-02); Dividend comparison engine (2026-08-02)

Boundary: Monotonicity is a computational property of otherwise identical fixed-envelope scenarios; it does not quantify how often a dynamic envelope will make additional export available.

No empirical pass is recorded. Computational verification is complete for the tested editions; measured-to-representative error remains blank until an eligible cohort is inspected.

Method: Inspected the observed-cohort eligibility and publication contracts, but no qualifying measured cohort receipt or public cohort output was available inside this sprint scope for comparison with representative model results.

Evidence: Observed cohort eligibility contract (2026-08-02); Observed cohort schema migration (2026-08-02)

Boundary: Golden cases, NASA-derived representative shapes and deterministic invariants are not measured external validation.

Matching fixtures passed; changed units failed; null and zero produced both value and missingness mismatches; a 28.36662032232715 raw value matched an explicitly declared 28.4 one-decimal display.

Method: Executed fixture-based checks for exact matching, unit mismatch, null-versus-zero, and declared display rounding through checkMetricCoherence and checkCoherenceManifest.

Evidence: Checker implementation (2026-08-02); Fixture tests (2026-08-02)

Boundary: The checker compares declared normalized meanings. It does not infer semantic equivalence between undocumented prose variants.

The raw 154.347 AUD/MWh value, NSW1/NEM geography, five-minute daily max-minus-min basis, AEMO-NEM-2026-07-31 vintage, Medium confidence and reported state matched; the visible 154 AUD/MWh value passed the declared zero-decimal display rule.

Method: Compared the July 2026 NSW1 price-spread value and provenance across the grid kernel mapping, production JSON, production CSV, page table formatting and reading rules.

Evidence: Production JSON metric (2026-08-02); Production CSV metric (2026-08-02); Human table (2026-08-02)

Boundary: This is one retained metric in one fixed monthly edition, not a claim that every Grid Edge row has been traced.

All five occurrences retained 1.5 kW as a site limit for Western Power single-phase basic EG on the large network without retailer off-take, with export-constraint-register-2026-08-01 and High source confidence where provenance is represented.

Method: Compared the Western Power 1.5 kW site fallback across the source register, public page and source receipt, JSON, CSV and public methodology.

Evidence: Source register row (2026-08-02); Production model JSON (2026-08-02); Public methodology (2026-08-02)

Boundary: The 5 kW ceiling remains a technical comparison bound, not a connection offer or interval-availability guarantee.

All traced local surfaces now identify the 127.03 AUD default result as interactive_model_only, canonicalBenchmarkPublished false and durableReleaseStatus provisional. Computational verification is passed on 2026-08-02 while empirical validation is pending with no cohort edition. The rounded 127 AUD card remains a reported interactive result, not a canonical benchmark.

Method: Compared the local schema-v2 interactive result and its publication object across the model core, JSON contract, CSV contract, JSON Schema, page and workbench copy, durable provisional release record and product contract.

Evidence: Schema-v2 model publication state (2026-08-02); CSV publication fields (2026-08-02); Closed JSON publication schema (2026-08-02); Visible publication boundary (2026-08-02)

Boundary: This pass verifies publication-state coherence and computational labeling. It does not publish a canonical benchmark or supply the pending measured cohort.

The 28.36662032232715% headline matched raw exports and the documented method, and the page's 28.4% display matched its one-decimal rule. Ausgrid remained null/not-reported in JSON, blank/not-reported in CSV and Not reported on the page; no zero was imputed.

Method: Rebuilt the fixed AER record, compared the all-DNSPs customer-share value across source-derived record, JSON, CSV, page and product method, and traced Ausgrid's unavailable duration through null, empty CSV and Not reported presentation.

Evidence: Fixed record builder (2026-08-02); Production JSON (2026-08-02); Product methodology (2026-08-02)

Boundary: The source-defined full-access headline remains an unweighted mean across positive reported values and is not customer weighted.

Export Flexibility now declares Medium output confidence with High regime and Medium solar confidence retained separately and a written transferability basis. Two-Way Network Performance v2 declares High source-fidelity confidence while explicitly excluding completeness of network reporting. Grid Edge and Two-Way Network Performance both have dedicated public methodology anchors and direct page links.

Method: Compared explicit output and source-fidelity confidence levels, their limitations, public methodology anchors and direct page links across the current local contracts.

Evidence: Two-way public record contract (2026-08-02); Export Flexibility source receipt (2026-08-02); Public methodology content (2026-08-02)

Boundary: These confidence labels describe the named evidence boundary. They do not convert computational checks into empirical validation or complete missing source reporting.