I halted the deployment after canary tests showed abnormal drift in out-of-sample risk metrics and increased error correlation across multiple critical sectors, indicating potential model instability. Stress scenarios demonstrated procyclical amplification under liquidity shocks and breached predefined loss-threshold guardrails, elevating the risk of systemic spillovers if released. Additionally, governance checks revealed incomplete documentation and an unresolved audit finding on data lineage, preventing full compliance sign‑off. Given these combined model-risk and compliance signals, a halt is the most prudent and reversible step while remediation and revalidation occur.