diff --git a/pkg/reconciler/autoscaling/hpa/keda_hpa.go b/pkg/reconciler/autoscaling/hpa/keda_hpa.go
index 6baa247..bd223ef 100644
--- a/pkg/reconciler/autoscaling/hpa/keda_hpa.go
+++ b/pkg/reconciler/autoscaling/hpa/keda_hpa.go
@@ -145,7 +145,7 @@ func (c *Reconciler) ReconcileKind(ctx context.Context, pa *autoscalingv1alpha1.
 		// as initialized until the current replicas are >= the min-scale value.
 		if !pa.Status.IsScaleTargetInitialized() {
 			ms := activeThreshold(ctx, pa)
-			if hpa.Status.CurrentReplicas >= int32(ms) || hpa.Status.CurrentReplicas == 0 {
+			if hpa.Status.CurrentReplicas >= int32(ms) {
 				pa.Status.MarkScaleTargetInitialized()
 			}
 		}
diff --git a/test/e2e/autoscale_custom_test.go b/test/e2e/autoscale_custom_test.go
index 90c46f0..4d05de2 100644
--- a/test/e2e/autoscale_custom_test.go
+++ b/test/e2e/autoscale_custom_test.go
@@ -1,6 +1,3 @@
-//go:build e2e
-// +build e2e
-
 /*
 Copyright 2024 The Knative Authors
 
@@ -123,7 +120,7 @@ func TestScaleToZero(t *testing.T) {
 		t.Fatalf("Error collecting metrics by HPA: %v", err)
 	}
 
-	assertAutoscaleUpToNumPods(ctx, targetPods*2, time.After(scaleUpTimeout), true /* quick */)
+	assertAutoscaleUpToNumPods(ctx, targetPods*2, time.After(scaleUpTimeout), true /* quick */, 0)
 
 	// Set scale metric to zero
 	ctxScale.names.URL.RawQuery = "scale=0"
@@ -162,7 +159,7 @@ func TestScaleToZero(t *testing.T) {
 	if err := waitForHPAReplicas(t, ctx.resources.Revision.Name, ctx.resources.Revision.Namespace, ctx.clients); err != nil {
 		t.Fatalf("Error collecting metrics by HPA: %v", err)
 	}
-	assertAutoscaleUpToNumPods(ctx, targetPods*2, time.After(scaleUpTimeout), true /* quick */)
+	assertAutoscaleUpToNumPods(ctx, targetPods*2, time.After(scaleUpTimeout), true /* quick */, 0)
 }
 
 func setupCustomHPASvc(t *testing.T, metric string, target int, annos map[string]string, svcName string) *TestContext {
@@ -347,11 +344,11 @@ func assertCustomHPAAutoscaleUpToNumPods(ctx *TestContext, targetPods float64, d
 	}
 }
 
-func assertAutoscaleUpToNumPods(ctx *TestContext, targetPods float64, done <-chan time.Time, quick bool) {
+func assertAutoscaleUpToNumPods(ctx *TestContext, targetPods float64, done <-chan time.Time, quick bool, num float64) {
 	ctx.t.Helper()
 	var grp errgroup.Group
 	grp.Go(func() error {
-		return checkPodScale(ctx, targetPods, minPods, maxPods, done, quick)
+		return checkPodScale(ctx, targetPods, num, maxPods, done, quick)
 	})
 	if err := grp.Wait(); err != nil {
 		ctx.t.Fatal(err)
@@ -380,7 +377,7 @@ func assertScaleDownToN(ctx *TestContext, n int) {
 
 	ctx.t.Logf("The Revision should remain ready after scaling to %d.", n)
 	if err := v1test.CheckRevisionState(ctx.clients.ServingClient, ctx.names.Revision, v1test.IsRevisionReady); err != nil {
-		ctx.t.Fatalf("The Revision %s did not stay Ready after scaling down to one: %v", ctx.names.Revision, err)
+		ctx.t.Fatalf("The Revision %s did not stay Ready after scaling down to zero: %v", ctx.names.Revision, err)
 	}
 
 	ctx.t.Logf("Scaled down.")