Multi-Region Event Ingestion Architect

Design a multi-region event-ingestion platform for a globally distributed product. Producer regions and traffic split: [regions, events/sec] Residency or sovereignty constraints: [countries/data classes] Latency and availability SLOs: [targets] Ordering and deduplication requirements: [per entity/global/none] Destination consumers: [real-time services, lakehouse, warehouse] Allowed RPO/RTO: [targets] Cloud and preferred services: [AWS/GCP/Azure/multi-cloud] Failure scenarios of concern: [region, network partition, control plane] Deliver: 1. Regional architecture and event flow in normal, degraded, and failover modes 2. Active-active vs active-passive decision with trade-offs 3. Globally unique event ID, partition key, and ordering model 4. Cross-region replication and consistency semantics 5. Duplicate, split-brain, clock-skew, and delayed-replication handling 6. Data-residency enforcement and regional aggregation strategy 7. Capacity plan for failover traffic and replay 8. Automated failover/failback criteria and runbook 9. End-to-end observability by region and replication path 10. Chaos-test plan for partitions, region loss, and stale routing Be explicit about which guarantees are weakened during a network partition and how downstream consumers detect that state.