This is definitely a really useful component to the interview, but I don't think it should be the only component. A lot of our jobs are doing straightforward things: "implement spec X, where X is pretty close to what's been done before." But occasionally we'll encounter a problem that requires us to think up an algorithm. It might be similar to what's out there, but it's not going to be a googleable solution. It's these cases that define a good engineer.

I'd liken our jobs to those of pilots. Most of the time, anyone can fly the plane: just punch in the flight plan to autopilot, and press 'A/P engage'. However, when a pilot is near cumulonimbus clouds with an engine failure, he is forced to step up to the plate. This is what pilots are paid for: to handle the crazy emergency situations that can arise when you're in the air. This, too, is what good engineers are paid for: handling the difficult problems that occasionally arise during development. If you're not hiring for this, you're hiring code monkeys.