1. Map the current path
Record the code source, build command, runtime, configuration, database steps and background services. Identify the live release and who is authorized to operate it.
How the work proceeds
Rehearse the release and the decision to stop.
Agree on a useful first change and the checks that establish its result.
2. Investigate the unreliable step
Use sanitized errors and timing to separate a build failure, deployment problem and application failure after release. Reproduce the relevant condition in a controlled environment.
3. Agree on the change and acceptance
Choose a bounded improvement. Name the task it must preserve, the failure it should reveal and the point where the release would be stopped.
4. Rehearse the sequence
Test the identified build, data steps and worker behavior. Check recovery with controlled records and document anything that cannot simply be undone.
5. Release and hand over
Use the agreed window and access. Verify the application task, record the released revision and checks, and leave instructions for follow-up and recovery.
Keep the evidence close to the release.
The record should explain what ran and what was checked. Use the release checklist.