docs: correct Attic postmortem monitoring findings (#288) ## Correction The original postmortem said Attic had no health check. Gatus already checked it every minute and worked: - first failure: 10:10:57 CDT - ntfy alert after three failures: 10:12:58 - recovery alert: 11:21:58 The response failed because one noisy monitoring stream made the important Attic alert look routine. Buildbot and voicemail also fired transient alerts at nearly the same time. ## Changes - correct duration and timeline - replace the false missing-monitor claim with alert fatigue - mark the DB URL audit and fixed endpoint checks complete - add bounded follow-ups to rank noisy checks and split alert urgency - state why Attic should not trigger automatic rollback Docs only. All pre-commit hooks pass. Co-authored-by: Ammar Nanjiani <ammar.nanjiani@gmail.com> Reviewed-on: https://codeberg.org/ananjiani/infra/pulls/288