You're forgetting about market share. If prices don't sink further because demand elasticity is maxed out it has to be the result of some form of market failure. Either on the seller side (collusion, aka refusal to use pricing for market share increase) or irrationality on the buyer side (veblen goods, clearly a factor with cars and moon travel is, in a way, off the charts in that metric).
I like to look at sinking prices of technology through the the mental model of an inherent minimum cost, where learning or scale doesn't just make it magically cheaper but reduces avoidable overhead. Plenty of product types that have started expensive and then became cheaper and cheaper have clearly seen all the economy of scale that could reasonably be expected, they wouldn't become any cheaper per unit if the market consumed 10x as much. Same for spending a few more decades making them, unlikely to learn ways to make it web cheaper.
>>If prices don't sink further because demand elasticity is maxed out it has to be the result of some form of market failure
The "normal/natural" state is that prices do not simply continue dropping always.
US automobile prices were at their best in the late 20s, with a model-T-costing <$10k adjusted. By that point annual sales were maxed, competition had taken most profits out of the game... Not incidentally, the depression was starting to happen.
In the industry, cars-as-fashion became a thing.
When an industry hits peak demand, prices don't just continue to fall until the industry becomes tiny. At least, that's not a very common pattern and the trend certainly doesn't just continue to here as a continuation of an expansionary learning curve.
Welfarist/marginalist economic models don't take into account that time/history/economics is directional. Expanding markets don't follow the same rules as contracting ones.
Replace "normal/natural" with "idealized": yes, free market theory promises quite strongly that if there is a more efficient way to produce something, eventually some contender will discover that way and begin pushing out competition until that competition adapts. That mechanism does not require demand elasticity at all, it can work just fine in a perfectly inelastic market. Food is the closest thing we have to an inelastic market and it's certainly not a market characterized by lack of price competition caused by that absence of demand elasticity.
Yes, expanding markets work different than stable or contacting markets. But that difference is that the expansion can gloss over certain failure modes. Yes, Intel might have occasionally pushed out a new Pentium generation if x86 had been the only CPU architecture in the world and AMD (and Cyrix) had not existed, because the demand for instructions per second rises with availability (this abstraction conveniently groups demand from replacement upgrades and demand from new use cases into a single bin). But that would have still been a broken market, even while the brokenness was temporarily hidden by demand elasticity/expansion.
Model T wasn't the most efficient moment of car manufacturing, not if you include features in the equation. A car with the performance of the model T wouldn't sell today for a dollar beyond scrap value, and even Tatas and Dacias can e.g. be started without cranking and so on. The closest equivalent to a Model T would be a Tuktuk and those are cheaper today even before adjusting for a full century of inflation.
I like to look at sinking prices of technology through the the mental model of an inherent minimum cost, where learning or scale doesn't just make it magically cheaper but reduces avoidable overhead. Plenty of product types that have started expensive and then became cheaper and cheaper have clearly seen all the economy of scale that could reasonably be expected, they wouldn't become any cheaper per unit if the market consumed 10x as much. Same for spending a few more decades making them, unlikely to learn ways to make it web cheaper.