To me, this is a different thing. I think most founders would say things that don't scale are hacks that let them go faster, but cause extra work on the back-end or a higher unit cost. The upshot is speed to revenue or product-market fit, and if you don't find that then you don't need to waste time building the thing. And these subsidies can go away eventually with product improvement.

The Mayo story seems like something else. While you might learn something by buying the product in retail once or twice, buying in bulk to simulate demand isn't an acceleration to learning something. It's this thought that if only people could find my product, or have an easy way to buy it, it will sell. But once you stop buying, the effect goes away. It reminds me more of a dilemma that larger companies talk about: juicing this quarter's earnings at the expense of next year's.

If this was done to juice metrics shown to investors, doesn’t this constitute fraud?

In a similar vein,

As a teen at my first job, we would prime our tip-jar with a dollar bill from the register in order to jump start tips.

You'd be amazed at how well that worked.