Managerial Meta-Agent Repairs for Confirmed Recurring Errors Should Trigger Review of Earlier Decisions

This position paper argues that when a worker appeal confirms a recurring error and prompts a managerial meta-agent repair, the organization should search for earlier decisions that may share the error. Repairing an approved shift swap that an old scheduling component recorded as an absence can improve later bonus recommendations while leaving earlier bonuses unchanged. The central question is who must find those earlier cases. The proposed update protocol has two outputs. The meta-agent proposes a repair for later decisions; versioned release tooling records the actual component changes; and a service that the updater cannot modify searches protected historical records and states why each case was referred for human review. A separate release-testing queue receives unexpected changes caused by the repair. A synthetic specification audit—a simulation that checks whether these routing rules behave as written—uses known affected cases, missing records, unavailable replay, and a deliberately inserted repair side effect. It makes the coverage and workload of both queues visible. Field studies remain necessary to measure workplace error frequency, legal and organizational feasibility, and completed remedies. The organization that owns the records and learns from the appeal carries the first search burden, while authorized people retain control of individual findings and pay corrections.