fix: treat local rate limit queueing separately

apps/api/internal/runner/retry_decision_test.go

@@ -93,6 +93,25 @@ func TestFailoverPolicyAllowsModelOverride(t *testing.T) {
 	}
 }
 
+func TestLocalRateLimitWaitsInQueueWithoutRetryOrFailover(t *testing.T) {
+	err := &localRateLimitError{
+		clientErr: &clients.ClientError{Code: "rate_limit", Message: "local capacity exceeded", Retryable: true},
+		cause:     &store.RateLimitExceededError{Metric: "concurrent", Retryable: true},
+	}
+	retryDecision := retryDecisionForCandidate(store.RuntimeModelCandidate{}, err)
+	if retryDecision.Retry || retryDecision.Reason != "local_rate_limit_wait_queue" {
+		t.Fatalf("local rate limit should not be same-client retry, got %+v", retryDecision)
+	}
+
+	failoverDecision := failoverDecisionForCandidate(store.RunnerPolicy{
+		Status:         "active",
+		FailoverPolicy: map[string]any{"enabled": true, "allowCategories": []any{"rate_limit"}},
+	}, store.RuntimeModelCandidate{}, err)
+	if failoverDecision.Retry || failoverDecision.Action != "queue" || failoverDecision.Reason != "local_rate_limit_wait_queue" {
+		t.Fatalf("local rate limit should wait in queue without failover, got %+v", failoverDecision)
+	}
+}
+
 func TestProviderAuthErrorsFailOverInsteadOfHardStop(t *testing.T) {
 	runnerPolicy := store.RunnerPolicy{
 		Status: "active",