It looks like it was some corruption in the BullMQ / Redis Cache. I deleted the Queue Records, and implemented an eviction policy on the AMR instance. This seems to have remedied the issue.