What is missing is an open source orchestrator that has a feature freeze and isn't Nomad or docker swarm.
I think services take me literally half an hour a month or so to deal with unless something major has changed, and a major K8s version upgrade where I roll all nodes is a few hours.
If people are deploying clusters and not touching them for a year+ then like any system you're going to end up with endless tech debt that takes "significant planning" to upgrade. I wouldn't do a distro upgrade between Ubuntu LTS releases without expecting a lot of work, in fact I'd probably just rebuild the server(s) using tool of choice.
Running Docker Swarm in production, can't really complain, at least for scales where you need a few steps up from a single node with Docker Compose, but not to the point where you'd need triple digits of nodes. I reckon that's most of the companies out there. The Compose specification is really simple and your ingress can be whatever web server you prefer configured as a reverse proxy.
At that point, is Nomad still simple? If you're going to take on all of the essential complexity of deploying software at scale, just do it right and use Kubernetes.
Source: running thousands of containers in production.
Kubernetes uses etcd for service discovery. It isn't that Nomad does things differently or less simply, it is just that they are more explicit about it.
The real difference is that Kubernetes has a wide array of cloud hosts that hide the complexity from users, whereas Nomad can realistically be self hosted
But I'd love LTS release chain that keeps config same for at least 2-3 years.