I'd argue that if you can't tell the difference between a good system design and a bad system design other than "see if it works," you shouldn't be in a position to be giving system design interviews. The idea isn't to get a working system, it's to see how the candidate approaches a system design. If it's a problem you know how to solve and you can't see problems with their approach, they're probably at least as good at it as you are.

My experience is that you might be able to fake up to a certain point. After that, the interviewer can keep throwing curve balls and see how the candidate will approach those. Of course, it all depends on the problem given. I’ve seen system design questions that are way too abstract to be useful but the best ones I’ve been asked were very real life scenarios.

Producing a working system during a 1-hour interview is not a good measure of skill. That's not what interviews are for.