Unfortunately, I yet lack the reputation to comment, so I have to put this in as an answer.
Even after your explanation:
There are alternative forms of roman numerals,
… I think that Tom is right and the supposed solution is at least very debatable. It would at least be rather obscure, and I could not find any statement in the article you mention that supports its validity.
While there are a lot of variants that may all be considered correct forms of their time instead of just typos, there is not a total freedom in how the letters are arranged. As far as I know, and according with the Wikipedia article, the following basic rules are never broken in any variant:
1) Roman numerals are read and added from left to right.
2) The values of the summands decrease from left to right.
3) There is a subtractive notation as a shorthand, in which a single (!) letter left from a greater letter is considered to be negative and is subtracted from the greater. (I'd also insist that only
C were allowed to be used for subtractive notation, but we don't need that part here.)
While it seems difficult to agree on a single set of historical rules, probably because there isn't one, not adhering to at least these 3 rules above would IMHO break the whole system and introduce a lot of ambiguity. For instance rule 3 (which you make use of) is possible only because of rules 1 and 2, otherwise you'd never know what to subtract. If you ignore 1 or 2, then rule 3 makes no sense.
So, from my understanding, 1449 would be noted as MCDIL.
If you insist that double subtraction is a valid thing, I'd be really curious to see where that has been officially used before, and how exactly (see next point). Not to doubt you, but to learn.
At the very least (and if we assume for the sake of argument that MILD is a valid, unambiguous numeral), it means 1451, not 1449, as you would need to subtract I from L first, and then the result from D.
The "closest" valid numeral to MILD, if I'm right, would be MDIL, meaning 1549. While 1451 would be MCDLI.
Apart from that, as you asked for tips on how to pose the question, so as for the 400
I'd not use 400 but 500 in the hint, as it is at least the closest multiple of 100, which helps more for verifying the answer than 400. The 400 here feels just as arbitrary to me as, say, 39.
and as for the phrasing
I'd not say that the number IS more than 500 years old, as that doesn't make any sense to me with just some random number that could resemble a year past. It would be OK with special numbers like e or π, which have been detected to have a special meaning and thus given a special name at some point in the past. And as you see this is what the solvers were going for first. I'd rather say the number LOOKS to be more than 500 years old.
Anyway, thanks for the puzzle.