diff --git a/vertical-pod-autoscaler/pkg/recommender/input/cluster_feeder.go b/vertical-pod-autoscaler/pkg/recommender/input/cluster_feeder.go
index b32094057fa5..e891da80f0e4 100644
--- a/vertical-pod-autoscaler/pkg/recommender/input/cluster_feeder.go
+++ b/vertical-pod-autoscaler/pkg/recommender/input/cluster_feeder.go
@@ -25,6 +25,7 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/fields"
 	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/util/wait"
 	"k8s.io/apimachinery/pkg/watch"
 	vpa_types "k8s.io/autoscaler/vertical-pod-autoscaler/pkg/apis/autoscaling.k8s.io/v1"
 	vpa_clientset "k8s.io/autoscaler/vertical-pod-autoscaler/pkg/client/clientset/versioned"
@@ -50,6 +51,8 @@ import (
 )
 
 const (
+	evictionWatchRetryWait                     = 10 * time.Second
+	evictionWatchJitterFactor                  = 0.5
 	scaleCacheLoopPeriod         time.Duration = 7 * time.Second
 	scaleCacheEntryLifetime      time.Duration = time.Hour
 	scaleCacheEntryFreshnessTime time.Duration = 10 * time.Minute
@@ -142,14 +145,21 @@ func WatchEvictionEventsWithRetries(kubeClient kube_client.Interface, observer o
 			FieldSelector: "reason=Evicted",
 		}
 
-		for {
+		watchEvictionEventsOnce := func() {
 			watchInterface, err := kubeClient.CoreV1().Events(namespace).Watch(context.TODO(), options)
 			if err != nil {
 				klog.Errorf("Cannot initialize watching events. Reason %v", err)
-				continue
+				return
 			}
 			watchEvictionEvents(watchInterface.ResultChan(), observer)
 		}
+		for {
+			watchEvictionEventsOnce()
+			// Wait between attempts, retrying too often breaks API server.
+			waitTime := wait.Jitter(evictionWatchRetryWait, evictionWatchJitterFactor)
+			klog.V(1).Infof("An attempt to watch eviction events finished. Waiting %v before the next one.", waitTime)
+			time.Sleep(waitTime)
+		}
 	}()
 }