Container Apps: Stuck Revisions in Activating in West US 3 with GPU workload profile

## Problem description

My Azure Container App in West US 3 cannot start using the Consumption-GPU-NC24-A100 workload profile. The control-plane Start operation reports Succeeded, but revision remains permanently in Activating with health state None. Azure creates a replacement replica approximately every 63 seconds. Each replica becomes NotRunning, reports zero containers, and produces no console logs. This indicates failure before container creation, not an application crash.

Environment

• Region: West US 3 Profile: Consumption-GPU-NC24-A100- CPU: 24- Memory: 220Gi- Minimum/maximum replicas: 1/1

Latest reproduction

• Start result: Succeeded- Actual revision state: Activating- Replica state: NotRunning- Container count: 0- Console result: Could not find container

Diagnostics completed

Azure diagnostics reported no image-pull, storage-mount, OCI runtime, quota, node-pool, KEDA, health-probe, or container-exit failures. The same Consumption-GPU-NC24-A100 workload shape with 24 CPU and 220Gi runs successfully in East US.

Requested assistance

Please investigate the internal GPU scheduler and host-allocation logs for the correlation ID above. Please confirm whether usable A100 capacity exists in West US 3 and whether there is a regional platform incident. The issue is unresolved and remains reproducible. Please keep the support request open and escalate it to the Azure Container Apps GPU platform team.