This problem occurs when the NoMachine cluster is configured to use the SSH protocol. A symptom of the problem is that the MAC/IP address mapping for the NoMachine failover cluster happens several times a minute.

The nxerror.log file reports messages like:
16198 17863 14:18:43 222.890 DaemonLogin/DaemonLogin: ERROR! Invalid reply
'nx> 103 hello on board nx.
nx> 106 nxserver-7.6.2 letse os' context[C].

As a temporary workaround, configure NoMachine cluster to use the default NX protocol. To do that, execute in a terminal on the primary server:

sudo /etc/NX/nxserver --clusterupdate --proto nx:4000