No, I have not failed to distinguis between niche and bigger market - and don't call it "consumer" because it need not be. Even industry, military, automotive are looking at COTS - but not COTS designed for "consumer" quality levels, as you might find when you buy the cheapest CD player or battery charger at an outlet.

If you have a product that you expect to sell 10 of, or maybe 10/year, you can normally buy extra components directly when you produce the first series - or when you get the last-order notification. And the reason is that a customer who buys such products knows he has to pay a lot to receive such specialized hardware. Some of the extra money to cover the development costs. And some to cover the availability costs.

The thing is that the market isn't driven by special-case products selling in tens/year. So component manufacturers can't spend money on second-sourcing components when 99% of the users of the components will reevaluate their production costs every 18 months and plans to jump ship if they can save $0.50 by using a different component (which does not imply "consumer" or "low-quality" or similar).

Richards post did say "any developer". And it really is not true that "any developer" would chose components based on second-sourcing, for the simple reason that the bulk of all production are not of products where it is meaningful to stick with a specific component or supplier for longer times.

It really does matter what the customer wants to pay for. The customer who wants to pay 5 times the price to get the same product delivered 5 or 10 years from now can get that. But the simplest solution to that problem is to use the huge difference between sell price and production price to stock up on all components so you can make more production runs at a later time. Why else charge 5 times more for the product?

Most customers instead tries to figure out what their "standard" needs are, so that they can have multiple COTS solutions for the backbone of their needs. They may then be able to live with a specific 4-20mA sensor going off the market. Because they can see most of their 4-20mA sensors as COTS products. And they can see the equipment that reads the value from the 4-20mA sensor and reports the values in as COTS products. As long as they get a standardized protocol for how to configure the equipment and what the returned values looks like.

In the end, the bulk of all products tries to standardize on multiple suppliers of a service or a protocol, allowing the product owner to not need to worry about multiple suppliers of specific components. So the majority of designers can focus on how to build a product that is economical to produce compared to the price the customer is willing to pay, and that has a quality level that is matching the quality level the customer wants to pay for. But the majority of developers don't need to consider any 10 year availability from multiple component manufacturers.

It is the ability to standardize on a service instead of on a component that have lead to the huge technological acceleration. Our embedded products can solve more and more problems, while consuming less power and requiring smaller, easier to mount, boxes. 20 years ago, a bigger PLD really was big. And you had to send someone out with the car to reconfigure it. Today, you have a crypto key to be allowed to access it and tweek the parameters on-the-fly. Most industries don't want to buy more of the 20 year old PLC. It's worth their money to replace them with new equipment. And for the situations where it's too much work to switch PLC, they get large loads of replaced PLC that they can use as spare parts. All worth it because of the savings they got from replacing other PLC.

In the end, we are talking about an economical driving force. That is a driving force that are not much effected by the "eddy currents" you get from a few products sold in tens. And so, it isn't a driving force that makes it meaningful for "any developer" to spend too much time requiring second-source availability.

Why do PIC processors sell? Why do AVR processors sell? Why do most ARM processors sell? Millions and millions of processors that doesn't have any second-source supplier. They are not used by developers who have failed to distinguis between niche and consumer. But by developers who knows that they can stock up if their product is niche. And they don't need second-source if their product isn't niche.

So nothing DRAMATICALLY about it, at all.

So haven't I been involved in products where critcal components have stopped being produced? Yes! Ever had thousands of components in stock to be able to produce after the component was dropped? Yes! Haven't that resulted in problems? Yes! Would it have been better to have had second-source supplies and just be able to continue with an unchanged product? No - in all cases have the redesigned product either drastically reduced the production costs, or added significant functionality improvements. Even for a product expected to sell in a couple of hundreds, a redesign can save large amounts of money. Or result in new markets that was originally not thought about.

You are basically complaining about me generalizing about mass-production. But you did fail to notice that my post was complaining about the generalizing that second-sourcing is a general need. That is a DRAMATICALL difference ;)