feat(db): expose connection startup timeline - #7286

Why

Pool outcomes and coarse pool-state metrics show pressure after the fact, but they do not show when a checkout began or which writer-connection safety premise failed. Operators need bounded evidence that distinguishes pool saturation from slow or failing connection construction during boot and later pool growth.

What

• Count the start of each instrumented database checkout and route the huddle-history lookup through the existing typed operation wrapper.
• Instrument writer-pool construction and every new writer connection across physical-connect success, created-at floor setup, session timeouts, isolation verification, and ready.
• Emit exact, secret-safe lifecycle receipts on the process boot sequence alongside closed-cardinality counters and duration histograms with frozen label vocabularies and series budgets.
• Cover the production SQLx hook, initial minimum connections, post-start pool growth, cancellation, setup/isolation failures, Prometheus export, and relay boot lifecycle.
• Document how to interpret the acquisition and writer-connection signals.

Risk Assessment

Medium — this changes the production writer-pool after_connect hook and adds one counter increment to instrumented checkout paths. The database safety operations and failure semantics are unchanged; telemetry uses fixed enums, excludes URLs/raw errors, and has an explicit per-pod series ceiling. The principal residual risk is observability overhead when SQLx creates replacement connections, bounded to synchronous fixed-schema records and fixed-cardinality metrics.

References

• Builds on the early-startup lifecycle foundation merged in feat(relay): add early startup lifecycle logs #7258; rebasing onto current main removed that duplicate commit from this PR.
• Staging validation used pre-rebase head f58e9480a4f068db0c591f604fd6800fdd4bfc45. All runtime and deployment files in this PR are byte-identical at the rebased head; the only tree difference between those heads is an unrelated NIP-FI document newly present on main.
• Multi-architecture staging image: manifest sha256:b9351fa644e08376cbe1999f9bee311d33d1799a68eadef7929c4f862a832fec from https://github.com/block/buzz/actions/runs/33780888255.
• Staging deployment: https://github.com/squareup/builderbot-platform-core-infrastructure/pull/317. Both pods in ReplicaSet buzz-6c8758bd7d became Ready with zero restarts, every observed initial writer connection reached db_ready, and the new metrics produced data.
• Dashboard: https://app.datadoghq.com/dashboard/tm7-qxr-wt2/buzz-startup--rollout-safety.
• Originating discussion: buzz://message?channel=6ac85131-70cd-4bda-a031-38d34114934e&id=fa2bed181c092697210a60bb6eedc55a665d5c1c6cabc1413a04686647011f71.

Generated with Codex