This problem has been reported on Windows and it is triggered by the fact that all socket handles get into an error state at the same time, for reasons that are unknown at the moment and that don't depend on NoMachine.

When these circumstances occur, NoMachine processes start using a whole CPU core and the logs files start growing indefinitely with the following lines being repeated:

_NXThreadRead: ERROR! Failed to read signals from FD#1340 with result -1 error 108.
_NXSlaveWait: ERROR! Unexpected error on the signaling pipe.

The infinite loop should be prevented and NoMachine processes should fail gracefully.