A Copilot cloud agent is working on a branch that changed the test harness. The team expected Copilot Memory to reuse a stored repository fact about database test setup, but the agent does not apply it. Which option best explains the behavior?

The correct answer is that the repository fact failed validation. The provided Copilot Memory lookup clearly states "validation: citation range no longer matches current branch" and "memory decision: ignored." This indicates that while the fact was a candidate, its associated citation (tests/README.mdL44-L52) was no longer valid on the copilot/fix-refund-retry branch, likely due to changes in the tests/README.md file. Copilot Memory prioritizes accuracy and will ignore facts if their supporting evidence is no longer verifiable in the current context. "Copilot Memory is disabled everywhere" is incorrect because the lookup process itself occurred, indicating memory is active. "The workflow trigger skipped the run" is irrelevant to why a specific memory fact was ignored during an active agent session. "An MCP server returned stale data" is a speculative infrastructure issue not supported by the explicit validation failure message.