Evidence and recovery
Telemetry becomes stale at three minutes and fails at five. Eligible tracked offers become stale after two hours. Provider failures and Qwen outages degrade the relevant capabilities without taking the website offline. Database snapshots and isolated restoration tests record separate timestamps. Documentation publication must match the exact app commit before activation.
Observe is not required for price collection or existing sessions, but new Google sign-ins require its broker. Logs, ZIP codes, query text, account details, and email delivery details remain private. See PriceDip operations for source-specific limitations and recovery.
Dedicated Qwen and private search reachability checks preserve degraded states while web availability stays independent. The dashboard includes request latency, email budgets, per-source failures, deployment and rollback timestamps. Failure simulations cover three/five-minute telemetry thresholds, stale products, documentation mismatch, Qwen outage, and recovery.
Workflow evidence covers the latest main-branch application validation, wiki publication, and daily documentation reconciliation. Failed runs are failures; pending, missing, or stale workflow evidence remains unknown. Wiki publications are recorded in deployment history. The daily fleet inspector imports PriceDip operational findings into its existing issue and incident workflow.