Match behaviour. Separate state.

Video transcript

1. 0:00–0:06

   Start with the journey • Name the behaviour the test must establish. • Booking page • B17 · test record • Q17 + W17 · test queue + worker • Controlled recipient • Invented journey · verify state, permissions and failure behaviour

   An invented booking journey connects the page to B17, a test record, then Q17 and W17, its test queue and worker, and finally a controlled notification recipient. The proposed expectation is one correctly owned fixture record with supported permissions, failure and retry behaviour. The animation explains the path to verify; it does not claim the test ran or passed. Match the relevant production behaviour while documenting smaller capacity, mocks and scheduled-job differences.

2. 0:06–0:12

   Follow every connection • Separate data, credentials, queues and recipients. • Test request • Test queue • Test worker • Live state • Test identities + test keys • Separate data + queues + recipients • Another URL does not isolate state

   The complete reviewed request, queue and worker diagram stays inside staging, with the attempted crossing to live state blocked. Native labels focus on test identities and keys, then separate data, queues and recipients. A distinct URL or source branch alone does not isolate resources. Inspect databases, files, caches, payment and notification credentials, worker configuration, callback destinations and analytics. Test identities should not have live powers. Protect staging with access controls: noindex is not authentication.

3. 0:12–0:18

   Use defined fixtures • Include failure, retry and supported languages. • B17 · repeated request • Inspect duplicate handling • D17 · dependency fails • Recoverable state; no false completion • AR/EN · mixed IDs • Correct language + direction • Defined start → actual evidence → reset only owned test records

   Three synthetic fixture cases connect to expected behaviour to inspect: B17 repeats a request and requires checking duplicate handling; D17 fails a dependency and should leave a recoverable state without a false completion or real charge; mixed Arabic, English and identifiers require correct language and direction. Add ordinary, missing-optional and expired-permission cases as relevant. Define the starting state, capture actual evidence and reset only owned test records; this is not permission for broad shared-data deletion or a claim of exactly-once delivery.

4. 0:18–0:24

   Record what was simulated • A sandbox response proves only its stated checks. • Sandbox request • Validation only • SendGrid sandbox: request checks only • No delivery / events / template-content validation • Controlled delivery • Rendered message • Separate authorised test · inbox placement is not guaranteed

   Two reviewed evidence tracks remain separate. SendGrid sandbox checks request structure, types and format but does not deliver email, produce delivery events or validate referenced template content. It therefore cannot establish rendered-message correctness. A separately authorised delivery test to a controlled recipient supports inspection of the rendered message; it does not guarantee inbox placement. Document each provider simulator and unsupported behaviour rather than treating every successful sandbox response as a live result.

5. 0:24–0:30

   Verify production too • Passing staging does not prove the live release. • S17 · staging evidence • P17 · agreed safe live check • Record fixtures, mocks and differences • Record actual candidate + journey evidence • Name drift, fixture, access and cleanup owners; verify current inclusions

   S17 staging evidence connects only to a record of its fixtures, mocks and differences. P17 connects separately to agreed safe checks of the actual production candidate and user journey. Passing staging does not prove live settings, external resources, capacity or the deployed release. Capture real observations without unintended customer messages, charges or unrelated state changes. Verify production indexing controls separately, and name owners for drift, fixture updates, access, cleanup and costs. Review current service inclusions before defining additional work.