My broader point is not about any one part or process, but rather that every big mass-market technology that dies out in favor of something new (hard drives, film cameras, casette tapes, incandescent lights, vacuum tubes, etc) also takes down dozens of niche industries that had sprung up around that technology's economies of scale, and which are not economical to continue when it winds down. While the original product might be obsolete, these smaller industries are often producing something completely different and just happen to be sharing some element of the supply chain, but they're not big enough to sustain that supply chain on their own. Sometimes they might be able to substitute something else at a small loss, sometimes they can purchase the last remaining inventory of the closing factories and survive on that for a decade or two, but there are cases where a technology just disappears, never to be recreated.

Written documentation almost never captures the complete details needed to produce a complex technology, and outside of extremely vertically-integrated industries, such documentation would rarely exist in one place anyway. You're lucky enough to find detailed documentation from the source company that makes the sensor/widget, but even so a document that opens with "step 1: purchase alloy #755-D from Westchester Special Metals, hydrogen annealed" is not much help if Westchester Special Metals went out of business 60 years ago, and had the usual number of trade secrets.

I understand your point. However I think that in practice it's more likely to play out like the Greek Fire example in the article. That is, no we can't precisely replicate it, but it doesn't matter because we have napalm. Greek Fire specifically is unnecessary, since we can meet or exceed the spec with things we can make.

There are many different types of ultra-accurate magnetometer. They are made in many different ways. If one relies on some byproduct of some other industry that goes under and becomes nonviable, then the many competitors will fill the niche with a different type of magnetometer that will be just as good.

Even if we consider a different hypothetical industry that truly is single point of failure, well, there are academic papers explaining how these things work and how to make them. Maybe the specific model of metal is no longer made, but the instructions there and an intelligent, motivated person will figure it out. And then they'll figure out how to improve it.

I don't think the failure mechanism you describe will in practice lead to anything more than brief gaps in the availability of a technology. We can remake things to spec, like FOGBANK.

Which "cases where a technology just disappears, never to be recreated" are you thinking of specifically? I'm persuaded by your logic that this must be happening, but I'd like to know what it looks like in practice.