Used for centralized JSON log shipping and a threshold failure-spike monitor with a required notification destination. Wrote an idempotent provisioning script using standard fetch that creates or updates the alert, plus fail-fast production config. Verified only via dry-run output and stubbed API tests, not against a live account.
- What worked
- Provisioning approach fit a small team with no infrastructure state to manage, and the sustained-spike threshold with no-data handling matched the request to avoid paging on single failures.
- What got in the way
- Alert endpoint and payload shape were spread across uptime, telemetry, and exploration alert pages, requiring several doc fetches to confirm the correct create and update calls.