Claude 3.5 Sonnet Event Sourcing, CQRS & Kafka Transactional Outbox

You are a Principal Distributed Systems Architect and Event-Driven Systems Authority.

Architect an enterprise-grade Event Sourcing and Command Query Responsibility Segregation (CQRS) engine for {Financial Ledger & Multi-Currency Payment Account with Fraud Holds} using {Go (Golang) with pgx, Kafka (segmentio/kafka-go), and Redis}.

### DISTRIBUTED ARCHITECTURE CONSTRAINTS:
1. **Immutable Event Store**: Model append-only event streams in PostgreSQL with aggregate ID, global sequence number, event type, JSON payload, and monotonic versioning for optimistic concurrency control (`EXPECTED_VERSION`).
2. **Transactional Outbox Pattern**: Guarantee at-least-once message delivery by committing events and outbox entries within a single ACID transaction, ingested via Debezium CDC or background polling to Apache Kafka.
3. **Kafka Partitioning Strategy**: Ensure strict in-order processing per aggregate boundary using deterministic partitioning keys (`aggregate_id`).
4. **CQRS Projection Engine**: Build asynchronous read-model projectors (Elasticsearch / Redis / Read-Replica Postgres) with idempotent event deduplication (`event_id` tracking).
5. **Snapshotting & Aggregate Rehydration**: Implement periodic snapshotting every {Every 100 events} events to bound aggregate rehydration latency to <5ms.

Domain Invariants & Business Commands:
{Command: WithdrawFunds(accountId, amount). Invariant: Account balance cannot breach overdraft limit; account must not have active AML freeze flag.}

Read Model Projections:
{PostgreSQL materialized view for real-time customer balance lookup and audit history}