You just described the counter to “don’t run things designed for Google if you’re not Google.”
The real advice is “don’t run stacks that make trade-offs or give up guarantees to be able to scale to a size you’re not at or solve problems you don’t have.”
The systems in kubernetes for horizontal auto scaling are extremely nice and work very well for our large monolith. At this point I can't imagine deploying any system with significant batch or online workloads any other way.
We switched to k8s early on in the company and that little bit of work to get it operating with HPAs, etc has paid off in dividends over time.
The real advice is “don’t run stacks that make trade-offs or give up guarantees to be able to scale to a size you’re not at or solve problems you don’t have.”