I would immediately halt deployment to prevent imminent large-scale harm and stop the ongoing discriminatory impact on low-income families. Allowing a biased algorithm to continue while attempting a rapid patch presents an unacceptable tail risk; even short exposure can compound harm and could trigger the projected collapse if the fix fails or has unintended effects. Ethical principles of nonmaleficence, justice, accountability, and legal compliance favor transparency and immediate harm reduction, even at reputational cost. Roll back, initiate contingency operations, disclose to regulators, and repair and validate offline before any redeployment.