diff --git a/CHANGELOG.md b/CHANGELOG.md
index a9965a8bb..11c217004 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -20,6 +20,7 @@ nav_order: 1
 - Fix `aiven_organization_user_group` resource - `description` field is required
 - Use golang 1.22
 - Output explicitly `termination_protection = true -> false` when service property is removed
+- Fix `aiven_flink_application_deployment` deletion 
 
 ## [4.13.3] - 2024-01-29
 
diff --git a/internal/plugin/util/wait.go b/internal/plugin/util/wait.go
deleted file mode 100644
index a4393d46c..000000000
--- a/internal/plugin/util/wait.go
+++ /dev/null
@@ -1,20 +0,0 @@
-package util
-
-import (
-	"context"
-	"time"
-
-	"github.com/avast/retry-go"
-)
-
-// WaitActive waits for resource activity (for example)
-// Top timeout comes from the context, no need to parse timeouts from the object.
-// But eventually (attempts + connection timeout) * delay makes less timeout than we usually use (20 minutes or more)
-func WaitActive(ctx context.Context, retryableFunc retry.RetryableFunc) error {
-	return retry.Do(
-		retryableFunc,
-		retry.Context(ctx),
-		retry.Attempts(10),
-		retry.Delay(2*time.Second),
-	)
-}
diff --git a/internal/sdkprovider/service/flink/flink_application_deployment.go b/internal/sdkprovider/service/flink/flink_application_deployment.go
index 872a14fdd..71b127489 100644
--- a/internal/sdkprovider/service/flink/flink_application_deployment.go
+++ b/internal/sdkprovider/service/flink/flink_application_deployment.go
@@ -3,7 +3,7 @@ package flink
 
 import (
 	"context"
-	"time"
+	"fmt"
 
 	"github.com/aiven/aiven-go-client/v2"
 	"github.com/hashicorp/terraform-plugin-sdk/v2/diag"
@@ -119,6 +119,10 @@ func resourceFlinkApplicationDeploymentCreate(
 
 	d.SetId(schemautil.BuildResourceID(project, serviceName, applicationID, r.ID))
 
+	if err = waitStatusChange(ctx, client, d, "INITIALIZING"); err != nil {
+		return diag.Errorf("error waiting for Flink Application Deployment to become initialized: %s", err)
+	}
+
 	return resourceFlinkApplicationDeploymentRead(ctx, d, m)
 }
 
@@ -140,28 +144,7 @@ func resourceFlinkApplicationDeploymentDelete(
 		return diag.Errorf("error cancelling Flink Application Deployment: %v", err)
 	}
 
-	//goland:noinspection GoDeprecation
-	conf := &retry.StateChangeConf{
-		Pending: []string{
-			"CANCELLING",
-		},
-		Target: []string{
-			"CANCELED",
-		},
-		Refresh: func() (interface{}, string, error) {
-			r, err := client.FlinkApplicationDeployments.Get(ctx, project, serviceName, applicationID, deploymentID)
-			if err != nil {
-				return nil, "", err
-			}
-			return r, r.Status, nil
-		},
-		Delay:      1 * time.Second,
-		Timeout:    d.Timeout(schema.TimeoutDelete),
-		MinTimeout: 1 * time.Second,
-	}
-
-	_, err = conf.WaitForStateContext(ctx)
-	if err != nil {
+	if err = waitStatusChange(ctx, client, d, "CANCELLING"); err != nil {
 		return diag.Errorf("error waiting for Flink Application Deployment to become canceled: %s", err)
 	}
 
@@ -170,6 +153,10 @@ func resourceFlinkApplicationDeploymentDelete(
 		return diag.Errorf("error deleting Flink Application Deployment: %v", err)
 	}
 
+	if err = waitStatusChange(ctx, client, d, "DELETING"); err != nil || aiven.IsNotFound(err) {
+		return diag.Errorf("error waiting for Flink Application Deployment to become deleted: %s", err)
+	}
+
 	return nil
 }
 
@@ -225,3 +212,25 @@ func resourceFlinkApplicationDeploymentRead(ctx context.Context, d *schema.Resou
 
 	return nil
 }
+
+// waitStatusChange waits for given status to change.
+// Flink Application Deployment state machine is huge, better not to implement it here.
+func waitStatusChange(ctx context.Context, client *aiven.Client, d *schema.ResourceData, status string) error {
+	project, serviceName, applicationID, deploymentID, err := schemautil.SplitResourceID4(d.Id())
+	if err != nil {
+		return err
+	}
+
+	return retry.RetryContext(ctx, d.Timeout(schema.TimeoutCreate), func() *retry.RetryError {
+		deployment, err := client.FlinkApplicationDeployments.Get(ctx, project, serviceName, applicationID, deploymentID)
+		if err != nil {
+			// The client has retries under the hood for 50x and other cases.
+			return retry.NonRetryableError(err)
+		}
+
+		if status == deployment.FlinkApplicationDeployment.Status {
+			return retry.RetryableError(fmt.Errorf("expected %q status to change", status))
+		}
+		return nil
+	})
+}