My command of quantitative metre is non existent, but to my knowledge a particular instance of α, ι, υ in a particular word was almost always either long or short: it was a property of the phonology of the word, and not an artefact of the metre.
The quantity of α, ι, υ in word roots is given in larger Ancient Greek dictionaries such as LSJ or DGE. If you scroll through, you will see entries where there are exceptions (hence the “almost” above), where one poet once will have used a different quantity for one of those vowels in the stem. Linguists to my knowledge have not treated that as metrical licence, but as linguistic variation: if a poet used the “wrong” length for a vowel, the assumption is that some speakers really were pronouncing it like that.
Again: that’s my outsider linguist impression. Specialists in metre may know better.