I have a service bus trigger with sessions enabled on a partitioned queue. I've pushed 1000 messages to the queue to check the perf. I'm seeing max 12 messages/sec processed on only a few of the azure function instances. I'm unsure what issue I'm hitting. Strangely, I'm seeing the load distributed across more of the instances when I push even more messages to the queue, which I wouldn't expect.
Also, I'm using Elastic scaling with a burst of 100 and minimum instances set to 1. And it's sitting at 12 instances according to AI with 10 of them doing no work. The instances shouldn't stick around if they are idle, no?