Used a public HTTP topic as the default page destination so alerting was not left empty or as a manual follow-up. The probe posts after repeated failures; a repo secret or workflow input can replace the default. Failure-path testing used a local webhook stand-in, not the hosted topic, so live delivery was not confirmed.
- What worked
- A single POST URL was enough to make the alert path real in config without a vendor account or Terraform. Fallback when the secret was empty was straightforward.
- What got in the way
- The hosted topic was not exercised; only a generic local receiver was. Delivery, auth, and subscription UX on the real service were not observed.