Lesson 10 — Release, Migration, and Incident Quality

Duration: 120 minutes

Included delivery track: Follow LAB_SETUP.md for the supplied files and local simulation. Required activities use these materials; connected deployments are optional extensions. Instructor debrief: answer key. Real OAuth, cloud releases and payments are simulated or discussed through evidence packets. Deck: 10-release-and-incident-quality
Lab: Ship/No-Ship Release Room

Learners will be able to

explain why a release candidate, migration, environment, and evidence must line up;
read a CI/release report for failed, skipped, advisory, and required checks;
make a calibrated ship, hold, or rollback recommendation; and
communicate an incident without inventing a root cause.

Instructor preparation

Reveal fixtures/release-rounds.md one round at a time. B27 and training:27 identify the same fictional candidate; training:26 is deliberately mismatched evidence. These labels are not cryptographic digests. No cloud or database action is required.

Agenda

Time	Facilitation
0–10	Arrival: ask, “What does a green dashboard fail to tell us?”
10–25	Teach promotion, immutable artifacts, migrations, environment parity, smoke tests, canary, rollback, and break-glass as distinct concepts.
25–40	Walk through the fictional release checklist: required checks fail closed, exact build evidence, synthetic data, and bounded production-safe checks.
40–50	Practice: classify CI states—passed, failed, skipped, advisory, cancelled, missing—and decide which blocks release.
50–60	Break.
60–95	Release room: teams receive new evidence in three rounds and update a recommendation after each round.
95–110	Incident debrief: write facts, impact, unknowns, containment, next evidence, and owner. Ban causal language not supported by evidence.
110–115	Discuss why no model or person gets to relabel a failed required check as passed.
115–120	Exit ticket: state the exact evidence you would need to promote a candidate.

Assessment

Use templates/release-recommendation.md. A learner passes by making a defensible decision; “ship” is not the preferred answer when evidence is incomplete.