This is incredibly common - I've heard a company end up rearchitecting their instance type choices due to Datadog billing on a per-node basis (with some peak usage billing shenanigans). Their business model unfortunately encourages some very specific architectures which doesn't work for everyone.