Regarding Metallb external IP allocation:
When a service of type LB gets allocated a specific IP (via the metallb.universe.tf/loadBalancerIPs annotation), and there are other LB services in the cluster, and metallb's controller restarts,
it will assign the specific IP to a different service.
Thus the request for specific IP fails.

4.15/latest

1. Deploy latest metallb.
2. Apply the following yamls:
apiVersion: metallb.io/v1beta1
kind: IPAddressPool
metadata:
  name: ipaddresspool
  namespace: metallb-system
spec:
  addresses:
  - 192.168.144.1/32
---
apiVersion: v1
kind: Service
metadata:
  annotations:
    metallb.universe.tf/loadBalancerIPs: 192.168.144.1
  name: nginx
  namespace: default
spec:
  ports:
  - nodePort: 32725
    port: 80
    protocol: TCP
    targetPort: 80
  type: LoadBalancer
---
apiVersion: v1
kind: Service
metadata:
  name: nginx1
  namespace: default
spec:
  ports:
  - nodePort: 32726
    port: 80
    protocol: TCP
    targetPort: 80
  type: LoadBalancer
---
apiVersion: v1
kind: Service
metadata:
  name: nginx2
  namespace: default
spec:
  ports:
  - nodePort: 32727
    port: 80
    protocol: TCP
    targetPort: 80
  type: LoadBalancer
3. See that the "nginx" service is assigned the address 192.168.144.1
4. Delete the controller pod. (may need to repeat more than once)
5. See that the address got assigned to a different service, leaving "nginx" pending.

The address 192.168.144.1 is assigned to a different service, and service "nginx" external-ip moved to pending.

The address 192.168.144.1 remains as the external ip of service "nginx"

Upstream issue: https://github.com/metallb/metallb/issues/1984