I think the point of the comment you replied to is that nobody builds prototypes in a scalable fashion, since if it was scalable (implying you spent some resource to make it scalable) it wouldn't be prototype.

The whole point of doing things that don't scale includes..... doing things that don't scale.

That includes things that can't scale.

That's the whole point.

Why discriminate against prototypes when we're doing things that don't scale and they're... well... things that don't scale? Making prototypes and testing assumptions with them is 100% in the spirit of the methodology.

I think there's some nuance about charging for the thing as if it were not a prototype - so the prototype part is (somewhat) hidden to the user. People just respond differently, and you get market testing.

Somehow in hardware this seems crazy and there's a million reasons not to actually sell the product (certification, hard to change in the field, it's expensive!), so I think it's done less often.

You can go for the other paradigm, however, which is more like hand-building each one, but could be thought of as a high-quality prototype meant for sale. It's much more expensive per unit, but you learn every time, and hey no tooling cost! Boosted Boards is the best example I can think of.