I think nobody has yet explained the origin of the construction. It would be an interesting thing to read, does anybody know that?
In my opinion, it's inevitable that most of our (human) categorizations produce (subjunctive? ) a certain number of monsters that are difficult to classify (when the categorization is not strict) or make up too small a category. (Those who are interested in mathematics may see a similarity with the classification of finite simple groups.) It's understandable that we want our categories as simple as possible but it's often impossible to achieve without giving up something else, perhaps as valuable.
I think that if we gave the "come March" sentence to a computer program taught some strict rules of recognizing word classes, we would quickly get the answer that come is a preposition there. It behaves like one and it doesn't behave like a verb. The only two possible verb hypotheses are the subjunctive and the imperative and both can be rejected because it's not the way the subjunctive or the imperative is used. And such an answer may be satisfactory.
There's still one problem though. This answer seems to be against some speakers' intuition (which was expressed in this thread). I understand that they "use it as a verb" by which I mean that they think it a verb when they use it. I, for one, have never used the construction, but whenever I read it I thought of come as of a verb, not a preposition. As I said, I'm not a linguist and I don't know whether it really is a problem. I read in this article that parts of speech are recognized by "the syntactic or morphological behaviour of the lexical item in question." So, it seems to me, users' feelings are irrelevant. Is it really so? What do you think?
Now, why didn't some members like the preposition hypothesis? I feel the answer must lie in the history of the construction. As I'm one of those who were displeased with the hypothesis at first, I'll try to justify my feelings.
I can't believe that it was first used this way with the intention of using a prepositional phrase. I believe it must have been originally used as a subjunctive clause. Perhaps it wasn't correct to use the subjunctive this way even then, but I believe this usage of come was never meant to be prepositional. It looks like the now-archaic use of the present subjunctive, if March be; if we substitute come for be and carry out the often-encountered if-inversion (wow, I have never produced so many hyphens un one sentence before!), we'll see exactly come March. That was my first thought when I started reading this thread. Do you think it could explain the origin of the construcion?
As for the users' feelings again, I recently read somewhere (probably on this forum) that let in let's dance was an auxiliary verb. It was stated as if it were a scientific fact and I believed it was. But now I think it's exactly a case of using one's feelings (created by knowledge of etymology) to determine the word class a word belongs to. I will have to resort to my native language now. Let's take an example of an English third-person imperative sentence:
Let him who is without sin cast the first stone.
Niech ten, który jest bez grzechu, rzuci pierwszy kamień.
The pair of sentences have a very nice feature: they mean exactly the same and they employ exactly the same word order. (The Polish sentence is one word shorter, which is because there are no articles in Polish. "The" is omitted.)
Let and niech both introduce the imperative mood. Him and ten are both subjects. Cast and rzuci are both something I don't know the English term for, but their function is clear. The rest is not important.
Since let and niech have exactly the same functions it would be natural to demand that they be in the same word class I think. But I have never heard of anybody saying that niech is an auxiliary verb. And I think the different approaches of Polish and English grammarians to exactly the same thing might be caused by the different feelings they have about the words let and niech respectively. Let is mainly a verb in English which, I guess, makes English grammarians call it a verb in this context too. Niech has no other function than this, so nothing makes Polish grammarians feel it's a verb. (Even though its origin is verbal. It happened too long ago to influence our thinking.)
I think, if my interpretation is correct, this could prove that linguists do not base entirely on syntax and morphology. If that's how it should be or not is another matter, which I would love to hear from you about.
This is the first time I have written such a long post on these forums! I still hope someone reads it though!