Application Gateway for Containers: Autoscaling Issue During Load Increase — Gateway Returning 5xx Responses

Problem description

I am experiencing an issue with Application Gateway for Containers (AGC) where, during a load increase from approximately 120 requests/sec to 2,600 requests/sec, the gateway begins returning gateway-generated HTTP 5xx responses. The backend services remain healthy, serving only 2xx responses, and the WAF logs show no blocks. The problem started on October 9, 2026, and reproduces during load tests. I am seeking insights into why the gateway is returning these errors under WAF inspection and how to mitigate this capacity limitation.

Environment

Application Gateway for Containers deployed in front of an Azure Kubernetes Service (AKS) workload, using delegated subnet dev-wus2-snet-agc in the /24 address space, with WAF policy attached.

What I've already tried

I have not tried anything yet, as I am seeking guidance on potential causes and troubleshooting steps.

Current status

I am looking for advice on whether AGC with WAF has minimum-instance or reserved-capacity settings, and how to diagnose potential capacity or scaling issues during load surges. I would also appreciate recommendations for further diagnostics or configuration adjustments.