[Nutanix]No host has enough available memory for VM, machine stuck in Provisioning and machineset scale/delete cannot delete machines

Nutanix machine without enough memory stuck in Provisioning and machineset scale/delete cannot work

Server Version: 
4.12.0
4.13.0-0.nightly-2023-01-17-152326

Always

1. Install Nutanix Cluster 
Template https://gitlab.cee.redhat.com/aosqe/flexy-templates/-/tree/master/functionality-testing/aos-4_12/ipi-on-nutanix//versioned-installer
master_num_memory: 32768
worker_num_memory: 16384
networkType: "OVNKubernetes"
installer_payload_image: quay.io/openshift-release-dev/ocp-release:4.12.0-x86_64 2.
3. Scale up the cluster worker machineset from 2 replicas to 40 replicas
4. Install a Infra machinesets with 3 replicas, and a Workload machinesets with 1 replica
Refer to this doc https://docs.openshift.com/container-platform/4.11/machine_management/creating-infrastructure-machinesets.html#machineset-yaml-nutanix_creating-infrastructure-machinesets  and config the following resource
VCPU=16
MEMORYMB=65536
MEMORYSIZE=64Gi

1. The new infra machines stuck in 'Provisioning' status for about 3 hours.

% oc get machines -A | grep Prov                                               
openshift-machine-api   qili-nut-big-jh468-infra-48mdt      Provisioning                                      175m
openshift-machine-api   qili-nut-big-jh468-infra-jnznv      Provisioning                                      175m
openshift-machine-api   qili-nut-big-jh468-infra-xp7xb      Provisioning                                      175m

2. Checking the Nutanix web console, I found 
infra machine 'qili-nut-big-jh468-infra-jnznv' had the following msg
"
No host has enough available memory for VM qili-nut-big-jh468-infra-48mdt (8d7eb6d6-a71e-4943-943a-397596f30db2) that uses 4 vCPUs and 65536MB of memory. You could try downsizing the VM, increasing host memory, power off some VMs, or moving the VM to a different host. Maximum allowable VM size is approximately 17921 MB
"

infra machine 'qili-nut-big-jh468-infra-jnznv' is not round

infra machine 'qili-nut-big-jh468-infra-xp7xb' is in green without warning.
But In must gather I found some error:
03:23:49openshift-machine-apinutanixcontrollerqili-nut-big-jh468-infra-xp7xbFailedCreateqili-nut-big-jh468-infra-xp7xb: reconciler failed to Create machine: failed to update machine with vm state: qili-nut-big-jh468-infra-xp7xb: failed to get node qili-nut-big-jh468-infra-xp7xb: Node "qili-nut-big-jh468-infra-xp7xb" not found

3. Scale down the worker machineset from 40 replicas to 30 replicas can not work. Still have 40 Running worker machines and 40 Ready nodes after about 3 hours.

% oc get machinesets -A
NAMESPACE               NAME                          DESIRED   CURRENT   READY   AVAILABLE   AGE
openshift-machine-api   qili-nut-big-jh468-infra      3         3                             176m
openshift-machine-api   qili-nut-big-jh468-worker     30        30        30      30          5h1m
openshift-machine-api   qili-nut-big-jh468-workload   1         1                             176m

% oc get machines -A | grep worker| grep Running -c
40

% oc get nodes | grep worker | grep Ready -c
40

4. I delete the infra machineset, but the machines still in Provisioning status and won't get deleted

% oc delete machineset -n openshift-machine-api   qili-nut-big-jh468-infra
machineset.machine.openshift.io "qili-nut-big-jh468-infra" deleted

% oc get machinesets -A
NAMESPACE               NAME                          DESIRED   CURRENT   READY   AVAILABLE   AGE
openshift-machine-api   qili-nut-big-jh468-worker     30        30        30      30          5h26m
openshift-machine-api   qili-nut-big-jh468-workload   1         1                             3h21m

% oc get machines -A | grep -v Running
NAMESPACE               NAME                                PHASE          TYPE   REGION    ZONE              AGE
openshift-machine-api   qili-nut-big-jh468-infra-48mdt      Provisioning                                      3h22m
openshift-machine-api   qili-nut-big-jh468-infra-jnznv      Provisioning                                      3h22m
openshift-machine-api   qili-nut-big-jh468-infra-xp7xb      Provisioning                                      3h22m
openshift-machine-api   qili-nut-big-jh468-workload-qdkvd                                                     3h22m

The new infra machines should be either Running or Failed.
Cluster worker machinest scaleup and down should not be impacted.

must-gather download url will be added to the comment.