My journey has been similar yet distinctly different. I went from "put it all on one server" to running servers in AWS. But the cost was preventing me from doing much more than run a couple of compute nodes. I hated the feeling of "I could setup a server to do X but it's gonna cost another $x/month". So I've been shifting back to my own servers.

I do like devops and automation though. Automation is brilliant for creating easily reproducible and stable environments - especially for things you don't touch very often.. Proxmox was what let me start moving back "on prem" as it were. There are "good enough" terraform plugins for proxmox that let me provision standardized VMs from a centralized code-base. And I've got ansible handling most of the setup/configure beyond that. I've now got like 20 VMs whereas before I only had 2 EC2 nodes due to cost. So much happier...

Biggest problem with this us actually outages/availability. Aws has 3-nine or 4-nine availability regardless and has several people ensuring it.