diff --git a/clustering/process.go b/clustering/process.go
index d0b1eeeb..7d9be5ee 100644
--- a/clustering/process.go
+++ b/clustering/process.go
@@ -218,7 +218,7 @@ func (p *managerProcess) do(ctx context.Context) (bool, error) {
 		return false, nil
 
 	case StateHealthy, StateDegraded:
-		if ss.NeedSwitch {
+		if ss.NeedSwitch && !ss.PreventPodDeletion {
 			if err := p.switchover(ctx, ss); err != nil {
 				event.SwitchOverFailed.Emit(ss.Cluster, p.recorder, err)
 				return false, fmt.Errorf("failed to switchover: %w", err)
diff --git a/e2e/failover_test.go b/e2e/failover_test.go
index cadd97f3..f3caee39 100644
--- a/e2e/failover_test.go
+++ b/e2e/failover_test.go
@@ -14,7 +14,7 @@ import (
 //go:embed testdata/failover.yaml
 var failoverYAML string
 
-var _ = Context("failure", Ordered, func() {
+var _ = Context("failover", Ordered, func() {
 	if doUpgrade {
 		return
 	}
diff --git a/e2e/prevent_delete_test.go b/e2e/prevent_delete_test.go
index cb211e33..3b497546 100644
--- a/e2e/prevent_delete_test.go
+++ b/e2e/prevent_delete_test.go
@@ -151,6 +151,7 @@ var _ = Context("PreventDelete", func() {
 			}
 			return nil
 		}).Should(Succeed())
+		time.Sleep(30 * time.Second)
 	})
 
 	It("should not finish rollout restart if replication delay occurs", func() {
@@ -206,6 +207,70 @@ var _ = Context("PreventDelete", func() {
 			return nil
 		}).Should(Succeed())
 
+		// wait for cluster to be healthy
+		Eventually(func() error {
+			cluster, err := getCluster("prevent-delete", "test")
+			Expect(err).NotTo(HaveOccurred())
+			for _, cond := range cluster.Status.Conditions {
+				if cond.Type != mocov1beta2.ConditionHealthy {
+					continue
+				}
+				if cond.Status == metav1.ConditionTrue {
+					return nil
+				}
+				return fmt.Errorf("cluster is not healthy: %s", cond.Status)
+			}
+			return errors.New("no health condition")
+		}).Should(Succeed())
+	})
+
+	It("should not finish switchover if replication delay occurs", func() {
+		cluster, err := getCluster("prevent-delete", "test")
+		Expect(err).NotTo(HaveOccurred())
+		primary := cluster.Status.CurrentPrimaryIndex
+
+		// set huge replication delay
+		setSourceDelay(0, 10000)
+
+		// wait for prevent-delete annotation to be added
+		Eventually(func() error {
+			out, err := kubectl(nil, "get", "pod", "-n", "prevent-delete", fmt.Sprintf("moco-test-%d", primary), "-o", "json")
+			Expect(err).NotTo(HaveOccurred())
+			pod := &corev1.Pod{}
+			err = json.Unmarshal(out, pod)
+			Expect(err).NotTo(HaveOccurred())
+			if val, exists := pod.Annotations[constants.AnnPreventDelete]; !exists {
+				return errors.New("annotation is not added")
+			} else if val != "true" {
+				return fmt.Errorf("annotation value is not true: %s", val)
+			}
+			return nil
+		}).Should(Succeed())
+
+		// never finish switchover
+		kubectlSafe(nil, "moco", "switchover", "-n", "prevent-delete", "test")
+		Consistently(func() error {
+			cluster, err := getCluster("prevent-delete", "test")
+			Expect(err).NotTo(HaveOccurred())
+			if cluster.Status.CurrentPrimaryIndex == primary {
+				return errors.New("switchover is not finished")
+			}
+			return nil
+		}, 1*time.Minute).ShouldNot(Succeed())
+
+		// resolve replication delay
+		setSourceDelay(0, 0)
+
+		// wait for switchover to be finished
+		Eventually(func() error {
+			cluster, err := getCluster("prevent-delete", "test")
+			Expect(err).NotTo(HaveOccurred())
+			if cluster.Status.CurrentPrimaryIndex == primary {
+				return errors.New("switchover is not finished")
+			}
+			return nil
+		}).Should(Succeed())
+
 		// wait for cluster to be healthy
 		Eventually(func() error {
 			cluster, err := getCluster("prevent-delete", "test")