Personally I think the amount of pain wasn't enough, so I'm using Canonical Maas, and I can just spin up extra servers, though going beyond 4 nodes seems kind of wasteful in a home setting. Though my rig seeps 100Watt, which for hobby purposes is okayish imo. I've been running this since Ubuntu 16.04 and upgrading to 18.04 has been a relative breeze compared to a situation where I would've built this using adhoc scripts(i.e. non-k8s setup.)
I did make a wireguard tunnel to a colocation datacenter for offsite backups using async mirroring of rook/ceph. B/c with so many moving parts and automation, wiping a cluster accidentally is quite possible.
Not that I dislike LXD, but a good configured server won't require many interventions. In my case, I was about to upgrade some guest SO, and I had to relearn lots of things and remember what I did, even why, after about two years without touching it.
Even if it’s just for me I still try to write reasonable commit messages and all that.