These arguments you describe have been made on Meta. Questions have been posted following the general form of:
Why was my question (link) deleted? I'm just asking about the coolest hidden features of x, and there are lots of other questions of that form, like (examples).
I know they exist because I have personally answered and/or voted to close them. I could probably go and dig up links, but there seems little purpose in doing so. Most of them have been closed as duplicates, and then later deleted. Only 10k+ users will be able to see them, and it really misses the point how many of those questions get asked.
The fact is, a far greater number of people come to Meta asking why they can no longer post questions of any kind. (Yeah, yeah, you wanted data, but this is plain as day, all you have to do is watch the recent questions page on Meta for a week; anyone who's a regular participant on here knows what I'm saying is true.) This is evidence that our automatic post ban system is working. People who post consistently low-quality content and show no indication of improving are no longer allowed to keep trying. It's the same "broken windows" logic, but applied in a slightly different way. The point is, the overabundance of those questions has not given anyone significant pause about the automatic post ban system. So obviously the real battle is not trying to figure out ways to deter people from asking stupid questions on Meta (and if it is, well, it's time to give up, because that ain't never gonna happen).
No, it seems to me that the point is making it incontrovertibly clear what our standards are on Stack Overflow. It's easy to close the aforementioned questions as duplicates of the canonical Q&A on the subject because it's completely clear what the violation was and what the current situation is. There's a clear guideline, you violated it repeatedly, and you pay the associated penalty. Like all rules, our goal is and should be consistency.
The argument people are making with regard to leaving the old, highly-valued but now off-topic questions on the site is that they are sending an unclear message about what our content standards are. The fact that people might come here to Meta and use them as examples is only a side-effect of the real problem. The real problem is that if we allow some of those questions, regardless of the date, it makes our decision to enforce the standing rules against new questions seem arbitrary and capricious.
The thing I don't understand is why the simple solution to this problem has been rejected by the "deletionists". That is, simply providing indicators that make it exceptionally clear that the viewed content is no longer allowed on Stack Overflow, but is retained only for historical purposes. To me, the beauty of this strategy is that it allows us to retain the content that one group wants to retain, while avoiding the harms cited by their opposition. In other words, it is a balance: exactly what one should be seeking in the case of seemingly deadlocked conflict.
If we provide multiple, exceptionally clear visual indications that an old question no longer meets the content guidelines for SO, then we avoid the problems described above of our enforcement of the rules seeming arbitrary and capricious, and the side-effect of people coming to Meta complaining that their new question (following the same model) was closed.
If and when people do come to Meta complaining that their question was closed and citing the historical exceptions as evidence, we can do what we at Meta do best: leave snarky comments and close their question as a duplicate of the canonical "What are these 'historically significant' questions, and why can't I ask new ones in the same vein?" question. You know, the same one that will be linked on the "historical significance" notice that appears on the page with each of the historically significant questions.
In other words, the people who might hypothetically make this argument if we retain these old questions will simply be wrong. They will have failed to read the guidelines that were clearly provided. We all know that we can't do anything to force people to read, but as long as this is all made clear and consistent, there can be no legitimate argument made and no one can be truly offended by what they perceive as an arbitrary enforcement of the rules.
I just don't understand what the other harms of leaving these questions on the site are. As I've said many times prior in comments: if you don't want to read these questions, you don't have to. They're not hurting your Internet, but they are hurting the Internet of people who might want to read them.
As long as we solved the "mixed signals" problem, we can just leave the questions alone, locked, not as broken windows, but as historical stained glass windows in a cathedral. You can look, but you can't touch.
I think some people, during the course of this discussion, have been either intentionally or inadvertently obscuring the distinction between those old, historically significant questions that actually have something useful to provide to the site, and those that are just garbage.
I see absolutely no point in keeping garbage around, but I think we should do everything within our power to keep potentially useful information alive and accessible on the Internet. I realize that this is incredibly subjective, but I think, like pornography, most people know it when they see it.
For example, there is a big difference, apparent to literally everyone, between The Definitive C++ Book Guide and List and Questions every good .NET developer should be able to answer? versus UNIX man page jokes? and Programmer's food. In other words, there's a glaringly obvious difference between those old potentially-bikeshed questions that actually provide a high volume of useful information, and those that are just purely nonsensical fun. Jeff seems to make the same argument here; I'm not sure how and why that got forgotten in the recent haste to bring the site up to ISO 9000 compliance.
And in the edge cases, we should err on the side of preservation, rather than obliteration. At least until we follow Shog9's advice and have a thorough discussion on the merits of each question here on Meta. Then, depending on the consensus, and only then, should we consider deleting the question for good.
We decided through experience that these types of questions are not generally a good fit for a Q&A site like Stack Overflow. They tend to attract poor quality answers and discussion, which are two things we obviously don't want to encourage. That's why we've modified our guidelines so as to no longer allow them. But these old questions are exceptions. (Or, at least, some of them are—the exceptional ones.) They managed to attract high-quality answers and accumulate lots of useful information from experts. They did this despite the general tendency of this type of question. But that's no reason to throw out the baby with the bathwater. That information already has been and likely will be useful to others in the future. On balance, taking the risk isn't worth it to allow these types of questions. But the risk has already been taken, so there's just no good reason not to preserve those where the outcome was, despite all odds, positive.