diff --git a/python/pyspark/ml/tests/connect/test_parity_classification.py b/python/pyspark/ml/tests/connect/test_parity_classification.py
index 7805546dba70..3c7e8ff71a2d 100644
--- a/python/pyspark/ml/tests/connect/test_parity_classification.py
+++ b/python/pyspark/ml/tests/connect/test_parity_classification.py
@@ -21,8 +21,6 @@
 from pyspark.testing.connectutils import ReusedConnectTestCase
 
 
-# TODO(SPARK-52764): Re-enable this test after fixing the flakiness.
-@unittest.skip("Disabled due to flakiness, should be enabled after fixing the issue")
 class ClassificationParityTests(ClassificationTestsMixin, ReusedConnectTestCase):
     pass
 
diff --git a/python/pyspark/ml/util.py b/python/pyspark/ml/util.py
index 5849f4a6dad8..f249b0070090 100644
--- a/python/pyspark/ml/util.py
+++ b/python/pyspark/ml/util.py
@@ -15,6 +15,7 @@
 # limitations under the License.
 #
 
+import asyncio
 import json
 import logging
 import os
@@ -154,12 +155,16 @@ def add_ref(self) -> None:
             self._ref_count += 1
 
     def release_ref(self) -> None:
+        should_del = False
         with self._lock:
             assert self._ref_count > 0
             self._ref_count -= 1
             if self._ref_count == 0:
-                # Delete the model if possible
-                del_remote_cache(self.ref_id)
+                should_del = True
+
+        if should_del:
+            # Delete the model if possible
+            asyncio.run(del_remote_cache(self.ref_id))
 
     def __str__(self) -> str:
         return self.ref_id
@@ -348,7 +353,7 @@ def remote_call() -> Any:
 
 
 # delete the object from the ml cache eagerly
-def del_remote_cache(ref_id: str) -> None:
+def del_remote_cache(ref_id: str):
     if ref_id is not None and "." not in ref_id:
         try:
             from pyspark.sql.connect.session import SparkSession