Briefly put, Roman numerals are an alternate form of enumeration. It makes it easy to see the difference between the foreword and the actual content of the book. This is also the reason it's used in nested lists.
The history is that page numbering and Arabic numerals started to become popular in English at about the same time. It wasn't until centuries later that prefaces began to be numbered like this.
History of page numbers
I found a source that explains the history really well:
Numbering pages started out not as a tool for readers but a guide for those who physically produced books. In Latin manuscripts copied in the British Isles as far back as the eighth or ninth century, numbering was sometimes used to ensure that individual sheets of parchment were collated in the correct order. Use of numbering was sparse. It’s been estimated that around 1450—just before the birth of printing in the West—less than 10 percent of manuscript books contained pagination.
Fifty years later, the proportion of now-printed works with pagination was much higher. Part of the change reflected the new role of page numbers. Rather than strictly being tools for compiling leaves in the proper order, by the 1510s scholars were starting to refer to page numbers of printed volumes in their own writing.
When Did Books Get Page Numbers—and Are They Even Useful Anymore?
Please note that this only applies to the "body" of the work. Also, a number of the earlier works had folio numbers, not page numbers. The preface material would usually not be numbered at all (it would sometimes be numbered with the body, other times it would have numbered sections). This book (1575), for example, has no numbering in the preface (except for what was penciled in later), although the rest of the volume is foliated.
It should also be noted that well into the 17th century some books still didn't have any page numbers.
Roman vs. Arabic
Roman numerals were the original number system used in Europe. Fibonacci is credited with popularizing Arabic numerals in Europe in 1202 with his book Liber Abaci. The book Numerical Notation: A Comparative History gives some more information about this.
English didn't see any Arabic numerals for quite some time, however. And it also took some time for the numerals to become popular and eventually replace Roman numerals in almost all contexts. Apparently, an important factor was the printing press, which created a new literate group unattached to the traditional Roman numerals.
The first time an Arabic numeral was used in an English book was William Caxton's book Reynard in 1481. It was a signature mark, "a2", which can be seen here at the bottom of the page. It was only the signature mark(s); the rest of the book used Roman numerals. Caxton published 6 books like this, until he went back to Roman numerals for the signature marks in 1484.
More and more books would have a mix of Roman and Arabic numerals starting in 1505 (where the date used Arabic numerals). The paper Numbering by the books: the transition from Roman to Arabic numerals in the early English printing tradition looked at all the books in the Early English Books Online database between the start of English printing and 1534 that contained both Arabic and Roman numerals (which was 55 books). Of those 55, there were 26 with Roman numeral foliation and 17 with Arabic numeral foliation. In 1523, the first book to only use Arabic numerals was published.
My own research found that only the earliest books (before 1600) used Roman numerals as page numbers (for the body). Prologues were either part of the body in terms of pagination or were not paginated at all.
Prefatory Material
The earliest instance I can find of book in English with separate pagination for the prologue and the rest of the book is Moderation truly stated (1704). The preface, entitled "A Prefatory Discourse" is over 50 pages, so it's convenient to have it be numbered.
Best Answer
While your argument for pluralizing the first noun is logical, in practice it doesn't seem to be a unanimously accepted standard. I can find examples of each style used in different places on the web. For some publications, there seems to be no standard at all:
The most definitive answer would be from a style guide; if you're publishing something, the publisher probably has a house style guide, and I believe legal departments of organizations often have style guides as well.
Here's what the King County style guide says, for example:
So it says to use the singular form "invoice" and to pluralize "Nos." It's typical that no especially compelling reasoning is given. Conformity is the main goal of this sort of organizational style guide, and logic comes second.
(There's no special reason for picking this particular guide, it was just easy to find. You don't need to follow its advice here; my point is that you should follow the advice given in your own particular style guide about this issue.)
By the way, it seems to me that there is a third option you didn't mention: "Questions Nos. 1, 2, and 3." Some people do in fact appear to take that option: "The Stone Fort And Manitoba Post Treaties Numbers 1 & 2." I also found one document that has an inconsistent use in the title versus in the text: Change List for Revision Numbers 1 and 2:
Of course, these are only a few examples and could easily be mistakes, but I just want to argue that this is a possibility.
In terms of logical arguments for one particular style: here's an argument for "Question Nos. 1, 2, and 3."
It's common in this kind of context to use "No. 1," "No. 2" etc. by themselves as noun phrases referring to objects rather than to actual numerals. And when they are used on their own as noun phrases, it seems logical to pluralize them when appropriate. For example, you can say "The questions were easy. Numbers 1 and 2 were multiple-choice, and numbers 3 and 4 were fill-in-the-blank."
Now suppose you want to clarify what type of objects Numbers 1 and 2 refer to. You can just add a noun adjunct before them, which will generally be in the singular, giving "Question Numbers 1 and 2." I don't know if this is the actual reasoning that led to this format, but hopefully this makes some sense.