Uneven memory distribution across worker replicas

Memory usage varies significantly across worker container replicas: most (60%) hover around 50% utilization, some (30%) at 20%, and outliers spike to 70%, while CPU remains consistently low (<10%) across all containers.

Investigate load balancing mechanism for BullMQ job distribution. Review worker configuration for consistency across replicas. Monitor individual worker queue depths to identify if specific workers are handling disproportionate workloads. Consider if job characteristics (payload size, complexity) cause uneven memory consumption.