diff --git a/pkg/scheduler/backend/queue/active_queue.go b/pkg/scheduler/backend/queue/active_queue.go
index 0fdd0f0d450..9581c65d996 100644
--- a/pkg/scheduler/backend/queue/active_queue.go
+++ b/pkg/scheduler/backend/queue/active_queue.go
@@ -324,14 +324,9 @@ func (aq *activeQueue) unlockedPop(logger klog.Logger) (*framework.QueuedPodInfo
 	aq.schedCycle++
 
 	// Update metrics for unschedulable plugins.
-	// Note: We don't clear UnschedulablePlugins and PendingPlugins here because:
-	// 1. If the pod schedules successfully, we need them for logging/debugging
-	// 2. If the pod fails to schedule, they will be cleared and repopulated in handleSchedulingFailure
 	for plugin := range pInfo.UnschedulablePlugins.Union(pInfo.PendingPlugins) {
 		metrics.UnschedulableReason(plugin, pInfo.Pod.Spec.SchedulerName).Dec()
 	}
-	pInfo.GatingPlugin = ""
-	pInfo.GatingPluginEvents = nil
 
 	return pInfo, nil
 }
diff --git a/pkg/scheduler/framework/types.go b/pkg/scheduler/framework/types.go
index 607f18ad683..6cb9e98afc3 100644
--- a/pkg/scheduler/framework/types.go
+++ b/pkg/scheduler/framework/types.go
@@ -619,6 +619,16 @@ func (pqi *QueuedPodInfo) DeepCopy() *QueuedPodInfo {
 	}
 }
 
+// ClearRejectorPlugins clears the plugin-related fields that track why a pod
+// was rejected in a previous scheduling attempt. This should be called at the
+// beginning of a new scheduling attempt to ensure stale data doesn't persist.
+func (pqi *QueuedPodInfo) ClearRejectorPlugins() {
+	pqi.UnschedulablePlugins.Clear()
+	pqi.PendingPlugins.Clear()
+	pqi.GatingPlugin = ""
+	pqi.GatingPluginEvents = nil
+}
+
 // PodInfo is a wrapper to a Pod with additional pre-computed information to
 // accelerate processing. This information is typically immutable (e.g., pre-processed
 // inter-pod affinity selectors).
diff --git a/pkg/scheduler/schedule_one.go b/pkg/scheduler/schedule_one.go
index 1160f456b8f..3a2fbe69ba7 100644
--- a/pkg/scheduler/schedule_one.go
+++ b/pkg/scheduler/schedule_one.go
@@ -1063,10 +1063,11 @@ func (sched *Scheduler) handleSchedulingFailure(ctx context.Context, fwk framewo
 	err := status.AsError()
 	errMsg := status.Message()
 
-	// Clear plugin sets to avoid stale data from previous scheduling attempts.
-	// They will be repopulated below for FitError cases.
-	podInfo.UnschedulablePlugins.Clear()
-	podInfo.PendingPlugins.Clear()
+	// Clear plugin-related fields to avoid stale data from previous scheduling attempts.
+	// These fields will be repopulated below for FitError cases.
+	// We clear them here (rather than at Pop) because we sometimes want to use them
+	// for logging when a pod schedules successfully (e.g., after being flushed).
+	podInfo.ClearRejectorPlugins()
 
 	if err == ErrNoNodesAvailable {
 		logger.V(2).Info("Unable to schedule pod; no nodes are registered to the cluster; waiting", "pod", klog.KObj(pod))