The patch deployed earlier that morning was rolled back to the previous stable version ( v2.4.1 ). This ensured that once the service restarted, it wouldn't repeat the same destructive logic.