The scenario emphasizes the need for immediate recovery of system stability in a production environment without retraining or rebuilding the model. This is a classic requirement for rollback capability , where operations can quickly revert to a previously validated and stable model version.

The correct lifecycle capability is redirecting production execution to a prior validated model state , which enables:

Rapid restoration of service continuity

Minimal operational disruption

Avoidance of time-consuming retraining or debugging during critical operations

Use of pre-approved, previously tested model versions

This capability is a core component of mature AI operations (MLOps), ensuring that organizations can manage risks associated with model updates.

Other options, while important, do not directly address the immediate need:

Controlled promotion paths ensure governance during deployment but do not enable instant rollback

Standardized metadata supports comparison and analysis but not real-time recovery

Lineage records ensure traceability and auditability but do not provide operational rollback capability

Although traceability is mentioned in the scenario, the primary requirement is fast recovery to a stable state , which is only achieved through rollback or version switching.

Therefore, the correct answer is Redirecting production execution to a prior validated model state , as it directly enables rapid recovery under operational constraints while maintaining governance and traceability.

=========