diff --git a/test/e2e/common/node/framework/podresize/resize.go b/test/e2e/common/node/framework/podresize/resize.go
index a9426587bac..5065883f62b 100644
--- a/test/e2e/common/node/framework/podresize/resize.go
+++ b/test/e2e/common/node/framework/podresize/resize.go
@@ -375,13 +375,14 @@ func WaitForPodResizeActuation(ctx context.Context, f *framework.Framework, podC
 					return "resize is infeasible"
 				}, nil
 			}
-			// TODO: Replace this check with a combination of checking the status.observedGeneration
-			// and the resize status when available.
-			if resourceErrs := VerifyPodStatusResources(pod, expectedContainers); resourceErrs != nil {
+
+			if pod.Status.ObservedGeneration < pod.Generation {
 				return func() string {
-					return fmt.Sprintf("container status resources don't match expected: %v", formatErrors(resourceErrs))
+					return fmt.Sprintf("waiting for observedGeneration (%d) to catch up to generation (%d)",
+						pod.Status.ObservedGeneration, pod.Generation)
 				}, nil
 			}
+
 			// Wait for kubelet to clear the resize status conditions.
 			for _, c := range pod.Status.Conditions {
 				if c.Type == v1.PodResizePending || c.Type == v1.PodResizeInProgress {
@@ -390,6 +391,7 @@ func WaitForPodResizeActuation(ctx context.Context, f *framework.Framework, podC
 					}, nil
 				}
 			}
+
 			// Wait for the pod to be ready.
 			if !podutils.IsPodReady(pod) {
 				return func() string { return "pod is not ready" }, nil
diff --git a/test/e2e/common/node/pod_level_resources_resize.go b/test/e2e/common/node/pod_level_resources_resize.go
index e4695f2dd76..a1e404516dc 100644
--- a/test/e2e/common/node/pod_level_resources_resize.go
+++ b/test/e2e/common/node/pod_level_resources_resize.go
@@ -71,17 +71,20 @@ func doGuaranteedPodLevelResizeTests(f *framework.Framework) {
 			expectedContainers := makeGuaranteedContainers(1, cpuPolicy, memPolicy, true, true, desiredCtrCPU, desiredCtrMem)
 			for i, c := range expectedContainers {
 				// If the pod has init containers, but we are not resizing them, keep the original resources.
-				if c.InitCtr && !resizeInitCtrs {
+				podLevelOnly := (desiredCtrCPU == "" && desiredPodCPU != originalCPU) || (desiredCtrMem == "" && desiredPodMem != originalMem)
+
+				if c.InitCtr && !resizeInitCtrs && !podLevelOnly {
 					c.Resources = originalContainers[i].Resources
 					expectedContainers[i] = c
 					continue
 				}
 				// For containers where the resize policy is "restart", we expect a restart.
 				expectRestart := int32(0)
-				if cpuPolicy == v1.RestartContainer && desiredCtrCPU != originalCtrCPU {
+
+				if cpuPolicy == v1.RestartContainer && (desiredCtrCPU != originalCtrCPU || (desiredCtrCPU == "" && desiredPodCPU != originalCPU)) {
 					expectRestart = 1
 				}
-				if memPolicy == v1.RestartContainer && desiredCtrMem != originalCtrMem {
+				if memPolicy == v1.RestartContainer && (desiredCtrMem != originalCtrMem || (desiredCtrMem == "" && desiredPodMem != originalMem)) {
 					expectRestart = 1
 				}
 				c.RestartCount = expectRestart