> and marginal costs either fall or remain constant
Not true. If your factory can make N widgets per year, and you want to make N+1 widgets, the marginal cost of the N+1th widget is vastly greater than the Nth widget.
I think the parent comment was talking about building factories and amortizing their costs over your unit production, whereas you assume the factory is a fixed cost with fixed capacity and looking at the marginal cost to produce a unit, which is really rare in many real world situations.
For most goods, the factory doesn't run anywhere near N, and the fixed costs are 6 or more orders of magnitude higher than the marginal widget costs, so your business is well served just by finding any method to use that plant more effectivly. As an example, I was quoted $60,000 for a mold which would have produced parts at $0.005. (Very small plastic widgets.) At that ratio, any amount of scale will increase my profit, since the marginal costs, even if they increase by a factor of 10 or 100, are negligible to the cost of the tooling. (And the global market for this widget is measured in hundreds of units.) Any amount of reusing the mold is going to save me money. Sure we have problems if we need N+1 widgets in less than 1/N more time, but if we expected to need 2N widgets, we could reuse the tooling design at a second factory, and marginal costs actually do keep dropping.
But in real life, by the time you received orders for N/2 widgets, you were already breaking ground on your next factory. And if you get an order for 100N, you smile because now you can switch to a higher productivity manufacturing method, like stamping instead of machining; at 10000N you can invest in mass-producing the machines-that-make-the-machines. This is how we end up being able to make even complex products like cars so cheaply that we have more cars than people.
Not true. If your factory can make N widgets per year, and you want to make N+1 widgets, the marginal cost of the N+1th widget is vastly greater than the Nth widget.