diff --git a/tensorrt_llm/executor/proxy.py b/tensorrt_llm/executor/proxy.py
index 11451bad5b67..6118561d534b 100644
--- a/tensorrt_llm/executor/proxy.py
+++ b/tensorrt_llm/executor/proxy.py
@@ -573,7 +573,9 @@ def mpi_done_callback(future: concurrent.futures.Future):
             raise RuntimeError(
                 "Executor worker returned error") from ready_signal
 
-        if isinstance(self.mpi_session, MpiPoolSession) and len(status) == 3:
+        # Use the polymorphic predicate: the module-level ``MpiPoolSession``
+        # symbol may be rebound to a factory at runtime, breaking ``isinstance``.
+        if not self.mpi_session.is_comm_session() and len(status) == 3:
             worker_process_identities: List[WorkerProcessIdentity] = status[2]
             self._worker_process_monitor.register(worker_process_identities)
 
diff --git a/tests/integration/test_lists/waives.txt b/tests/integration/test_lists/waives.txt
index e6fd0e0f0d8b..d44de6cf636e 100644
--- a/tests/integration/test_lists/waives.txt
+++ b/tests/integration/test_lists/waives.txt
@@ -89,7 +89,6 @@ accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_eagle3_guided_decoding_4gpus[
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_1gpu[v1_kv_cache-True-True-triton-auto] SKIP (https://nvbugs/6026676)
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[v1_kv_cache-dp4-cutlass-auto] SKIP (https://nvbugs/6388142)
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[v1_kv_cache-ep4-cutlass-auto] SKIP (https://nvbugs/5596343)
-accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[v1_kv_cache-ep4-trtllm-fp8] SKIP (https://nvbugs/6450338)
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[v1_kv_cache-tp4-cutlass-auto] SKIP (https://nvbugs/5596343)
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[v1_kv_cache-tp4-cutlass-fp8] SKIP (https://nvbugs/5651865)
 accuracy/test_llm_api_pytorch.py::TestGPTOSS::test_w4_4gpus[v2_kv_cache-dp4-cutlass-auto] SKIP (https://nvbugs/5596343)