Prior to the 2018.2.x release, the recommendation was to create a "Backup Primary" node for the situation you describe. So that if your Primary failed since it was the only node that could run the Licensing component you could failover to the Backup Primary to avoid any workers becoming unlicensed. Now with 2018.2, there's the "initial node" which runs Licensing, but if it fails you can now move the Licensing to *any* node in the cluster (there's no longer a Backup Primary).
OK, this is better than I thought.
Now that I know what I’m looking for, I found the documentation
Still, it would have been nice to be able to set up redundancy and auto-failover.
Is there no plan to make this happen automatically? Requiring manual intervention in a clustered deployment to maintain availability eliminates the purpose of the highly-available deployment in the first place.
Baby steps, right? I call this "high-ish availability".
For starters, license servers are evil. And to punish the user with total loss of production because the license server is offline is cruel and unusual. These are all factors that should count when selecting software, but of course we don't select it, we're just the ones who keep it running.