{
  "date": "2026-10-27",
  "requirement": "Complete explicit Order item removal and Order status transitions in all seven runtimes; preserve School, existing Order receipts and pending-release markers.",
  "scope": "Two remaining Order P1 cases plus a rerun of all eleven previous Order cases per runtime, using unchanged Order v1 and fixture.",
  "result": "91/91 native cases, including 14/14 new cells; SQLite 7/7, 75 independent database checkpoints per runtime (525 total). Prior concurrency coverage rerun: 49 two-process rounds, 98 workers.",
  "model": "One new evaluation: 0 errors, 1 accepted nullable-note warning, 2 optional suggestions. No model changes. cargo-teaql 2.0.14 verified; seven generated application guides and 92 scoped Assist responses retained in discovery/manifest.json.",
  "removalAcceptance": "Three isolated controls: save a note through a Notebook-only projection without deleting omitted items; explicitly mark SKU-PEN for audited soft deletion and recompute the complete bounded order total from 3850 to 2500; reject deletion from an old item version after a competing winner, rolling back the earlier order UPDATE and retaining every winner field. Fresh normal reads hide the tombstone; deleted-only reads return SKU-PEN at version -2. Independent SQLite checks retain all rows and compare every field, identity, relationship and unaffected timestamp.",
  "stateAcceptance": "Twelve sequential probes cover all nine DRAFT/CONFIRMED/CANCELLED pairs, an unknown target, and actual draft-edit calls against CONFIRMED and CANCELLED orders. Only DRAFT -> CONFIRMED, DRAFT -> CANCELLED and CONFIRMED -> CANCELLED commit. Repeated states reject without writing. Each rejection requires no provider UPDATE and an exact unchanged snapshot; each accepted transition increments the original version and changes only status and update timestamp. Fresh-process reads and independent SQL comparisons follow every probe. This matrix is application policy, not automatic KSML enforcement.",
  "nativeRepairs": [
    {"runtime":"rust","kind":"application compile repair","detail":"Three expression getters consumed their wrapper. Recreate the OrderItem expression for each getter; no generated or runtime source change. Failed compiler log retained."},
    {"runtime":"swift","kind":"application compile repair","detail":"Add try inside the two throwing map closures in complete-read; no generated or runtime source change. Failed compiler log retained."},
    {"runtime":"java","kind":"application transaction boundary repair","detail":"The initial graph save rejected stale deletion before issuing the order UPDATE, so it could not prove rollback of an earlier write. Use the published native TransactionExecutor from DefaultUserContext's runtime registry to save the fully reloaded scalar order first, then the marked item in the same provider transaction. The first 12/13 attempt and SQL ordering are retained. Final logs prove both writes and rollback. Generated APIs came from Assist; transaction APIs were inspected only in runtime source."},
    {"runtime":"typescript","kind":"existing runtime fix pending release","detail":"Published 0.3.1 produced the correct SQLite -2 tombstone but returned [] from deletedRowsOnly; the first 12/13 failure is explicitly indexed in the new receipt. Rebuild with --typescript-runtime ../teaql-ts using the existing School tombstone-read fix. Final workspace run passes 13/13. No new runtime patch was needed."}
  ],
  "transactionBoundary": "Java and Rust use explicit native provider transactions for removal, saving order then item. TypeScript, Python, Go, .NET and Swift save one loaded graph. All seven must prove the order UPDATE precedes the stale item failure and the order rolls back exactly; busy, locked and timeout errors never count as conflicts.",
  "runtimeQualification": "Java, Rust, Python, .NET and Swift pass with published runtimes. Go retains its existing workspace runtime and TypeScript now needs its existing School workspace fix for this expanded Order suite; both remain WORKSPACE PASS — RELEASE PENDING. Exact source fingerprints, commits, resolved packages, build and verification identities are in the receipt.",
  "generatorQualification": "Existing exact-model generator supplement remains pending deployment: TypeScript nullable scalar generation, generated console guides and Rust nullable update Assist. No generated domain-library source was inspected for discovery or edited manually in this stage.",
  "preservedEvidence": "The Order 56/56 and 77/77 receipts and their bound sources remain unchanged. Their narrower TypeScript published-runtime results do not cover the new tombstone query. School 238/238 and all five pending-release markers remain unchanged.",
  "mechanicalReview": "Source-bound native builds and SQLite runs, original-case reruns, scoped Assist identity, audited mutations, bounded query intent, raw-log hashes, fresh-process results, exact database deltas and historical receipt fingerprints are checked. Test results and documentation build are reported separately by the completion turn, not fabricated in this execution receipt.",
  "humanReviewFocus": "Business policy: repeated transitions reject; only DRAFT permits editing; CANCELLED is terminal. This fixture has no payment, refund, stock or tenant authorization effects. Future business changes require new acceptance evidence.",
  "telemetry": {"tokenUsage":"not reported","estimatedContextAvoided":"not measured","queryPolicyExecutionCount":"not reported","writePolicyExecutionCount":"not reported"},
  "limitations": "All Order P0/P1 coverage is complete for the recorded SQLite combinations. P2, HTTP, browser/TFP, ecosystem integrations, other providers, sustained stress, provider/commit faults, payment, stock and trusted tenancy still require their own evidence. Repeated controls and rounds do not inflate the 13-case count."
}
