Platform, SRE, and Operations playbook

Platform, SRE, and operations teams protect the production boundary. They decide how evidence may be collected safely, when a controlled change is authorized, what rollback is required, and how service health will be proven after action.

• Referenced command: the guide explains a command and its purpose.
• Externally executed: an authorized operator ran it outside OGO.
• Result captured: evidence from the command was committed to the path.
• Controlled change: an approved state change was executed with rollback and validation.

These states must not be collapsed. Referencing a command is not execution, and execution is not proof of improvement.

Before change, confirm scope, authorization, blast radius, rollback, monitoring, stop conditions, and the exact signal expected to improve. Avoid broad changes when a smaller evidence-producing action is available.

Validation should match the original impact and include system health, workload behavior, user experience, and an observation window appropriate to the failure mode. Record partial recovery honestly. If one path recovers while another remains degraded, the incident state is not fully resolved.

Change record, health proof, rollback status, return-to-service decision, residual risk, and follow-up owner.