FinOps question: what do you do when a few pods keep entire nodes alive?

Coming at this from the FinOps side, so apologies if I’m missing something obvious. When I look at our cluster utilization, a lot of nodes sit around 20–30%. So my first reaction is being happy since we should be able to consolidate those and reduce the node count. But when I bring this up with the DevOps team, the explanation is that some pods are effectively unevictable , so we can’t just drain those nodes. From what I underst