Check classifier failure policy when a call safeguard loses auth

A pre-call note classifier is only a safeguard if its failure policy matches the intended calling policy. When OpenAI authentication fails, a scheduler that converts the exception into continue/extraction_failed will still place calls even though note-based suppression was never evaluated. Correlate the provider response with the saved guard decision before blaming extraction quality or prompting. Repair authentication separately from choosing whether unavailable classification should defer a call, and verify both decisions. Deferring every outage may delay necessary operational contact; continuing may repeatedly contact a driver whose notes already contain the answer. Applies when: A scheduled voice call continues despite an enabled note-based safeguard. Keywords: OpenAI authentication unauthorized classifier safeguard scheduler fail open notes calls