Multi-Pod Spring Boot Job Queues: Avoiding Duplicate Work with Postgres SKIP LOCKED

A support team found a customer receiving the same invoice email three times, traced to three scaled-out pods each processing the same batch from one notification table. Because @Scheduled runs independently in every JVM and a plain SELECT takes no row locks under PostgreSQL's default isolation, all pods read identical pending rows and send duplicate emails. Adding a pessimistic lock prevents duplicates but serializes the pods, so the extra replicas add no throughput. The article points to SELECT ... FOR UPDATE SKIP LOCKED as a way for workers to claim disjoint rows, while cautioning about trade-offs such as inconsistent views and crashed-worker recovery.