In general I agree. However, there is still the case of labelling (e.g. safety symbols) that are applied directly to the product - an informational requirement that wouldn't make sense if we assume the above.
Symbols on labelling are a little different because they behave more like alarms - meaning, they do tell that something is happening, but the person has to act or inact.
(aside: I have some serious reservations regarding the efficacy/rationale of many required symbols. Similar to your explanation of "The crazy EU deviation", I suspect that such requirements were put forth by "someone with no [usability] engineering background"....but a topic for another thread, I suppose.
)
He's a lawyer.
Also, the symbols by themselves are not effective if the manufacturer does not "validate"their use in their medical device thru usability testing or the like.
But back to the topic at hand...
I still think the analogy holds for certain manufacturing contexts. Whether or not the information is intended to be read once (in the case of information for safety on product labelling), or several times (e.g. in the case of an assembly work instruction), in either case once competency is achieved the documentation is no longer an active control. But to the extent that we agree that such documented information can contribute to the competency in the first place, it is a valuable tool for controlling final outcomes, no?
It is, in particular because it has other uses (such as keeping the information for a new worker, facilitating audit, etc.).
But what seems to be the cern os this discussion is the need for documentation in general, not only for one use (such as enabling competency).
And the thing is, people in general are still too much focused on equating a quality system with a system of documents, when in reality is that que quality system is people working (I usually say, it's the worker pushing the button, not "only" the documentation the documentation that says he must push the button, if any).
I understand that this "interpretation" has several reason, being the historical focus, particularly from evaluators, in documentation, but what is important to understand is that documentation should be a defined requirement for any process, iF required, not always required per se, meaning, you should CONCLUDE if documentation is required, not presuppose that it always is.