It's a half truth. You can be sitting at a plateau in terms of how the game is percieved until you add a key piece of tech or a particular asset. Ideally, you can prove 100% of the concept in a preproduction state, where everything is functional but placeholder. That is not hard to achieve at small scopes. Sometimes you end up needing to go a little farther, though, as the aesthetics do make a tangible difference. It takes good design skills to know how well developed the concept is, and for really tech-driven stuff you have to try building it to understand it.

That said, it also depends on what elements are critical to the design goals. If you're driven by writing, you had better already have a novel or two of drafts already written before you sink any budget into it. If you're going for a tight set of action mechanics, expect to spend a long time looking at a prototype with a bunch of colored shapes.

There are many stories of wasted work in AAA, and a lot of it stems from being too hasty to pull the trigger on production and expecting the people making work at final quality to pull rabbits out of hats and fill in a lot of missing blanks as they go along, only to be sent back to the start for a retake of entire scenes.

Eeeeeh, "final" is such a strong word. They won't (or shouldn't) spend much time polishing assets before they've sufficiently tested out (and polished) the core gameplay. But yeah, they shouldn't invest too much until they're confident that the game's going in the right direction.

The Stanley Parable? I mean, without the script and voice acting, it would seem pretty dull.

Point and click adventure games seem like a good example - the mechanics for these are usually a pretty small part of what makes them fun. Awesome and polished art, sound, plot etc. are what matter.

Dark Souls, maybe. Without the polish I can't imagine it would be more frustrating than fun.