Address review comments: add documentation and TODOs

- Add documentation for CONFIG_MAP_NAME env var in prometheus.go
  explaining how it's set by Helm chart for multi-instance support
- Add TODO for scaleTargetRef.name CRD requirement in variant.go
  noting the check should be enforced at schema level

Related issues:
- llm-d#454: Consider hardware type when inferring target deployment
- llm-d#455: Replace pod filtering with app-level readiness probes

Author: clubanderson

internal/config/prometheus.go

@@ -20,7 +20,12 @@ const (
 	DefaultConfigMapName = "workload-variant-autoscaler-variantautoscaling-config"
 )
 
-// GetConfigMapName returns the ConfigMap name from environment variable or default
+// GetConfigMapName returns the ConfigMap name from environment variable or default.
+// The CONFIG_MAP_NAME environment variable is set by the Helm chart during deployment
+// (see charts/workload-variant-autoscaler/templates/manager/wva-deployment-controller-manager.yaml).
+// Each WVA deployment gets its own uniquely-named ConfigMap based on the Helm release name,
+// allowing multiple WVA instances to coexist in the same cluster without conflicts.
+// The Helm template sets this to: {{ include "workload-variant-autoscaler.fullname" . }}-variantautoscaling-config
 func GetConfigMapName() string {
 	if name := os.Getenv("CONFIG_MAP_NAME"); name != "" {
 		return name

internal/utils/variant.go

@@ -101,6 +101,9 @@ func filterVariantsByDeployment(ctx context.Context, client client.Client, filte
 		}
 
 		// Skip VAs without scaleTargetRef (required to know which deployment to look up)
+		// TODO: Remove this check once scaleTargetRef.name is made a required field in the CRD.
+		// This defensive check exists because the CRD currently allows empty scaleTargetRef,
+		// but it should be enforced at the schema level instead.
 		if va.Spec.ScaleTargetRef.Name == "" {
 			ctrl.LoggerFrom(ctx).V(logging.DEBUG).Info("Skipping VA without scaleTargetRef", "namespace", va.Namespace, "name", va.Name)
 			continue