In distributed systems, "just put a lock around it" sounds reasonable until you add retries, multiple regions, and a job that can run twice. Under real load you get duplicate processing, stuck workflows, and race conditions that only appear when everything is slow and concurrent.

In this session I'll show you how to use Cosmos DB as a coordination mechanism for distributed workflows, from basic lock/lease and leader-style patterns to saga orchestration for long-running work. We’ll focus on the correctness tradeoffs teams run into in production including contention, hotspots, retries, and how consistency choices affect what your system can safely assume across regions.