Currently, if the etcd container managed by podman-etcd is abruptly terminated, the monitor operation returns OCF_NOT_RUNNING. This is received by Pacemaker as the resource was never running, which triggers an immediate, local restart of the agent.

This restart is too quick and uncoordinated with the peer node. The agent attempts to rejoin a cluster that hasn't yet recognized the failure,
leading to inconsistent state detection (e.g., seeing 2 active nodes) and causing the start operation to fail or deadlock.