Reference TestingGuide
Two-laboratory reference Sandbox conformance and evidence tiers
Stage-level acceptance criteria for a genuinely joined future reference journey, without relabeling separate passing tests.
On this page
Current availability — October 25, 2026
Limitation
Reference Testing is NOT available in Production and is NOT operational in Sandbox. The relationship, Program-binding and order-link database packages are staged and unapplied, both reference gates are off, no native installation registry exists and no clinical mapping authority has been established. The Control Plane shows "Reference Testing configuration pending" and every action is disabled with its reason.
Everything below describes engineering source contracts and isolated test evidence (disposable PostgreSQL, loopback PostgREST, test-signed tokens). Where a behavior is described as "designed", it exists in source and isolated tests only — not as a joined, persisted canonical R1 workflow. This is not an availability, certification or publication announcement.
Current checked and unchecked evidence
Reference Testing is NOT operational in Sandbox. The owner-supplied independent selected baseline has 1,864 passing tests in 103 files, including 16 P4 source checks. This is selected source/isolated verification at the supplied baseline, not every historical suite, hosted browser acceptance or live vendor certification.
- CHECKED: P1 29 disposable PostgreSQL/PostgREST relationship/invitation cases, using locally signed test claims.
- CHECKED: P2 draft binding/privacy and P3 ten persisted refusal cases in disposable PostgreSQL/PostgREST.
- CHECKED: seven distinct blocked laboratory Control Plane workspaces, source/navigation assertions.
- UNCHECKED: genuine hosted A/B sign-in, invitation fragment claim through signup/six steps and browser state changes.
- UNCHECKED: same R1 canonical order→transmission→specimen→actual accession→released FINAL→CORRECTED→sender GET. R1-13 remains PARTIAL.
Prepare isolated actors and versioned prerequisites
- Use disposable fixtures for A sender, B performer and unrelated C; add B→C to prove dual-role identity without another account.
- Approve the A→B relationship with two independent actors and current version/effective window.
- Provision separate approved canonical Programs, clients and grants for A and C; never share buyer credentials.
- Obtain installed clinically reviewed mapping and native registry authority through a separately approved package: current P2 cannot provide them.
- Pin source/target fingerprints, specimen/collection/method/region, commercial terms, signer policy and installation environment.
Future joined positive acceptance sequence
NON_EXECUTABLE acceptance assertions
A authenticated command → one canonical order id O
O → one authorized transmission T (same Program / performer)
O → specimens S1,S2 → B receipt → actual accession X
O → B authorized official FINAL version 1
O → B authorized CORRECTED version 2; preserve version 1
A current scoped GET → version 2 + authorized history
C wrong Program / tenant GET → denied, zero report leakage
Revoke relationship → new referral refused; historic policy recheckedMandatory negative matrix
- Wrong tenant, sender/performer direction, environment, client, Program, installation or native code.
- Expired credential/grant, paused/revoked relationship, stale consent/Program/mapping version or fingerprint.
- Panel role/specimen/method/unit/region mismatch, withdrawn/future test, missing reviewer/commercial/technical authority.
- Same key changed payload conflict; 409/timeout no resend; unsigned and replayed callbacks.
- Receipt without accession, partial rejection leaving S2 unchanged, correction before FINAL and missing clinical release.
- Private cost, blocked code/specimen/price and cross-buyer result/webhook leakage all absent.
Evidence retention and release decision
A genuine joined test records stage-level database read-back for the SAME order with exact scope, versions, timestamps, clinical release provenance and official report references. Independent source unit tests plus a different canonical API journey are not a joined R1 proof. Test-signed JWTs and SET ROLE are not genuine Auth; CI success does not authorize Production. Capture final source revision and zero skips; unresolved authorities remain blockers.