Change data capture, dual-write patterns, eventual consistency, and keeping distributed financial data in sync across systems — the patterns that survive partial failures.
The dual-write problem and how to solve it: change data capture with Debezium, the transactional outbox pattern, idempotent downstreams, schema drift, and the reconciliation that catches the edge cases.
Active/active vs active/passive multi-region data, conflict resolution, and the trade-offs between latency, consistency, and operational simplicity.
Daily reconciliation jobs that catch CDC drift, the audit trail regulators want, and how to wire up automated break investigation.