This still is a theoretical assumption. A demo is the only proof to convince potential customers. I know it from experience. I made some interior tests for several game engines and there are huge differences when you really want to have a smooth and constant FPS rate. The same counts for large outdoors.

Currently I am creating a tech demo for a big game engine. I am under NDA and cant talk about details. But they use a sample world like this not only to show it to customers. They also use it early for testing new features, to see how the engine runs on the aimed platform, to see how texture size, fill rate issues, CPU or GPU can be a limiting factor. Thus you can test (as an example) if it is better to use the co-processor for mathematical operations or not. You can test dynamic or static batching, you can see how much an index buffer for vertices can speed things up, and so on.

A demo is more than just eye-candy, it is a technical playground for programmers and testers. It is important and has nothing to do with this strange debate of programmer's engine vs. art engine. Such thing does not exist. Programmer's and artists always have to work together, otherwise the project will fail.

The best example is supercan, it shows fantastic artwork and there must be good programmers behind it also.

And the workflow debate should not be about artists or programmers, it should be about speed and efficiency, about the fact to deliver in time, to spend your time to the important things.