HA VM remains stopped after migration on NFS v4 shared storage

The goal is to confirm whether a Proxmox VE HA-managed VM can reliably transition to the running state after live migration when its disk resides on an NFS v4 export, given that the same configuration succeeds in a single‑node test but fails in a multi‑node production cluster.

The uncertainty stems from the HA resource agent’s reliance on flock‑based locking, which may not behave as expected under certain NFS v4 server settings, and it is not documented whether the agent automatically detects such lock limitations and falls back to an alternative fencing mechanism.

Does the HA resource agent currently probe the NFS server for lock capability before attempting to start a VM?
Is there a configurable option to enable automatic fallback to a different locking or fencing strategy when lock acquisition fails on NFS v4?
Should administrators treat NFS v4 as unsuitable for HA workloads unless specific export options are applied, or can the cluster be made safe without manual storage changes?