The most common homelab mistake I see — and the one I've made several times — is adding too many services too fast. You discover Proxmox, you deploy Pi-hole, it works, you get excited, and three weekends later you have 14 LXC containers, half a dozen docker-compose stacks, a Kubernetes cluster you wrote a blog post about, and two databases nobody is backing up.
Then something breaks. You don't remember which container runs the thing. The Pi-hole you actually care about is down because the docker host went sideways updating an unrelated service. Your spouse can't get on the internet. You're fixing your homelab at midnight with a glass of wine and some regret.
The rule I now live by: one new service a month. Deploy it, use it for a few weeks, write down how it's configured, make sure backups work, and only then consider adding the next thing. If something breaks along the way, fix it before moving on — technical debt compounds faster in a homelab than it does anywhere else.
Also: back up your config. Proxmox snapshots, docker-compose files in git, secrets in Bitwarden. If your host dies and you can't rebuild in a weekend, your services are more fragile than they should be.
None of that means don't experiment. It means experiment with a plan. That's what separates a homelab that grows with you from one that collapses under its own weight.