Docker changes the network interfaces a fair amount, which is included in the calculation of the unique hash to define each swarm client. This means that if you use docker networking in your jobs, your swarm nodes will not keep the same name if they disconnect + reconnect. Disabling the unique ID makes it so they can't reconnect with the same name if a job was already running.

Desirable behavior is that the node goes down, reconnects under the same name, and is able to resume the jobs it had (assuming it has a resumable job, such as a pipeline job). Current behavior leaves you with lots of "stranded" nodes that are permanently offline and reconnected under a new name due to docker changing the network interfaces.