I'm in an SRE-like position for an otherwise unmanaged Kubernetes cluster. We have code merged in the repo as well as a few PRs still open.

I should note that by regularly I mean edge cases and quirkiness, not that your run of the mill cluster mostly using deployments is going to randomly explode. But our teams still find ways to regularly break it in unexpected ways.

So I guess you could call that anecdotal if you want.