Define production lol. I've used it for a CI cluster for a few years. Have had to recreate it a few times due to database corruption (despite using etcd across 3 nodes).
Honestly Rancher management is more of a pain than manually managing via ansible or something. And swapping to CRI-O backend instead of containerd js a huge pain for Rancher/RKE2, but pretty easy with k3s.
Disclaimer: I'm just a K3s hobbyist.
It really depends on your needs. At the companies I've worked for, they require some sort of support and guaranteed security, usually in the form of a contract. I do recall a note in the docs about using an external DB for HA, so probably check that out. Finally, what's the availability and resiliency requirements? That may impact your decision. Finally, you may be able to try it out in a nonproduction environment and save some dough while evaluating it's viability.
I've used truenas scale (comes with k3s preinstalled) and I ended up fucking up the cluster on accident a couple of times (mostly because of power outages), I've since moved over to Debian and a manual kubeadm install. Will make upgrading the cluster a bit more painful, but so far it's stable (ish, I have had issues with the Nvidia containerd backend)