If you run multiple instances, each one will execute the scheduled job, causing duplicates. Avoid it with leader election, a distributed lock (carefully), a single dedicated scheduler instance, or moving scheduling to an external system.