Design a scalable, highly available inventory management system for an e-commerce platform. Requirements: track stock across multiple warehouses; support Reserve, Release, and Purchase operations; prevent overselling under concurrent requests; provide at-least-once order events and idempotent APIs; expose read endpoints for product availability and per-warehouse breakdown; support flash-sale spikes (e.g., 50k RPS reads, 5k RPS writes). Specify: overall architecture (services, databases, caches, message buses), data model (items, warehouses, reservations, versioning), consistency model (strong vs. eventual) and why, strategies to avoid race conditions (optimistic concurrency/version numbers, conditional updates, idempotency keys, quorum or distributed locks where needed), scaling strategies (read replicas, sharding, write queues, CQRS), cache design and invalidation, handling of partial failures and retries, exactly-once semantics at boundaries, backpressure and rate limiting, observability (metrics, tracing), disaster recovery, and multi-region considerations.