Azure AI Gateway token metrics unavailable after configuring Application Insights telemetry

Question

I am configuring monitoring for an Azure API Management AI Gateway in Sweden Central.

Configuration

• AI Gateway workspace: default
• Model asset: coding-model
• Backend model deployment: gpt-5.6-luna
• Telemetry destination: Existing Azure Application Insights resource
• Application Insights resource: appi-opencode-swc-dev-001
• OTLP support: The portal indicated it would be enabled automatically
• Content safety policy: Lenient filtering for all configured categories
• Token rate limit: 10,000 tokens per minute per caller IP
• Request rate limit: 100 calls per 60 seconds per caller IP

The AI Gateway and model deployment are operational, but the monitoring blade displays:

Token metrics could not be loaded.

The portal returns an HTTP 400 error while querying the Prometheus-backed Azure

Monitor telemetry endpoint. Subscription IDs, resource IDs, monitoring account IDs, request IDs, and the full query URL have been redacted.

The affected metric is: azure.ai_gateway.client.token.usage

The issue appears limited to monitoring and does not prevent model requests from being processed.

Could anyone help confirm:

1. Whether this is a known provisioning issue with AI Gateway (preview) monitoring and an existing Application Insights resource.
2. Whether the managed Azure Monitor workspace and supporting resources were provisioned correctly.
3. Whether additional resources or RBAC permissions are required for OTLP support.
4. Whether monitoring should be reconfigured using a new Application Insights resource.
5. Whether the missing monitoring-account reference can be repaired by Microsoft.
6. Whether this AI Gateway monitoring capability is fully supported in Sweden Central.