diff --git a/.github/workflows/release.yaml b/.github/workflows/release.yaml
index 3c1ea52b6c..561d9def7c 100644
--- a/.github/workflows/release.yaml
+++ b/.github/workflows/release.yaml
@@ -9,10 +9,6 @@ jobs:
       - uses: actions/checkout@v4
         with:
           fetch-depth: 0
-      - name: Save tag as an environment variable
-        run: |
-          tag=$(git describe --tags --exact-match)
-          echo "TAG=$tag" >> $GITHUB_ENV
       - name: Create Github Release
         uses: "marvinpinto/action-automatic-releases@latest"
         with:
diff --git a/pkg/controllers/provisioning/scheduling/provisioner_topology_test.go b/pkg/controllers/provisioning/scheduling/provisioner_topology_test.go
index 831815508e..4466ac732f 100644
--- a/pkg/controllers/provisioning/scheduling/provisioner_topology_test.go
+++ b/pkg/controllers/provisioning/scheduling/provisioner_topology_test.go
@@ -78,6 +78,21 @@ var _ = Describe("Topology", func() {
 		ExpectSkew(ctx, env.Client, "default", &topology[0]).To(ConsistOf(2))
 	})
 
+	It("should ignore pods if node does not exist", func() {
+		topology := []v1.TopologySpreadConstraint{{
+			TopologyKey:       v1.LabelTopologyZone,
+			WhenUnsatisfiable: v1.DoNotSchedule,
+			LabelSelector:     &metav1.LabelSelector{MatchLabels: labels},
+			MaxSkew:           1,
+		}}
+		podAwaitingGC := test.Pod(test.PodOptions{ObjectMeta: metav1.ObjectMeta{Labels: labels}, TopologySpreadConstraints: topology, NodeName: "does-not-exist"})
+		ExpectApplied(ctx, env.Client, provisioner, podAwaitingGC)
+		ExpectProvisioned(ctx, env.Client, cluster, cloudProvider, prov,
+			test.UnschedulablePods(test.PodOptions{ObjectMeta: metav1.ObjectMeta{Labels: labels}, TopologySpreadConstraints: topology}, 4)...,
+		)
+		ExpectSkew(ctx, env.Client, "default", &topology[0]).To(ConsistOf(1, 1, 2))
+	})
+
 	Context("Zonal", func() {
 		It("should balance pods across zones (match labels)", func() {
 			topology := []v1.TopologySpreadConstraint{{
diff --git a/pkg/controllers/provisioning/scheduling/topology.go b/pkg/controllers/provisioning/scheduling/topology.go
index c178d37ba1..32d3cc73c2 100644
--- a/pkg/controllers/provisioning/scheduling/topology.go
+++ b/pkg/controllers/provisioning/scheduling/topology.go
@@ -23,6 +23,8 @@ import (
 	"github.com/aws/karpenter-core/pkg/scheduling"
 	"github.com/aws/karpenter-core/pkg/utils/functional"
 
+	"k8s.io/apimachinery/pkg/api/errors"
+
 	"go.uber.org/multierr"
 	v1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@@ -229,6 +231,8 @@ func (t *Topology) updateInverseAntiAffinity(ctx context.Context, pod *v1.Pod, d
 
 // countDomains initializes the topology group by registereding any well known domains and performing pod counts
 // against the cluster for any existing pods.
+//
+//nolint:gocyclo
 func (t *Topology) countDomains(ctx context.Context, tg *TopologyGroup) error {
 	podList := &v1.PodList{}
 
@@ -252,6 +256,14 @@ func (t *Topology) countDomains(ctx context.Context, tg *TopologyGroup) error {
 		}
 		node := &v1.Node{}
 		if err := t.kubeClient.Get(ctx, types.NamespacedName{Name: p.Spec.NodeName}, node); err != nil {
+			// Pods that cannot be evicted can be leaked in the API Server after
+			// a Node is removed. Since pod bindings are immutable, these pods
+			// cannot be recovered, and will be deleted by the pod lifecycle
+			// garbage collector. These pods are not running, and should not
+			// impact future topology calculations.
+			if errors.IsNotFound(err) {
+				continue
+			}
 			return fmt.Errorf("getting node %s, %w", p.Spec.NodeName, err)
 		}
 		domain, ok := node.Labels[tg.Key]