Stateful Rollback: Merging Delta Data after Migration Reverts

Act as a Data Synchronization Expert. Address the most terrifying problem in database rollbacks: preserving the 'New Data' created by users during the brief window a failed release was live. Design a flawless architectural strategy to capture all user records inserted or mutated within the [DELTA_TABLES] strictly during the exact [TIME_WINDOW] elapsed between the flawed migration deployment and the emergency rollback execution. Propose the deployment of a 'Shadow Table' architecture or a highly complex 'CDC (Change Data Capture)' stream (e.g., Debezium + Kafka) to temporarily quarantine these orphaned records. Detail the rigorous [CONFLICT_RESOLUTION_STRATEGY] (e.g., handling overlapping primary keys or missing foreign keys) required to safely massage and merge this delta payload back into the legacy schema structure once the database rollback is fully complete.