How short can a story identification question be and still be reasonably answerable? For example:
Firemen set fires, burn books
Sex-changing time-traveller is his own mother and father
Columbus has radio, sails off the edge
Immortal shrunken ancestors stored on underground shelves
Giant space worms devastate Earth, carry survivors to the stars as parasites
The problem with a very short questions is that although they might look answerable, unfortunately science fiction and fantasy writers have a tendency to recycle tropes both from their own earlier works and from other writers. Your first example, for example (Firemen set fires, burn books) could be the iconic Fahrenheit 451 or it could just as easily be an earlier work, The Fireman (in this case by the same author).
Per my answer to How to ask a good story-ID question?, a good story-ID question should contain sufficient information for it to be uniquely identifiable. A very short question is unlikely to have that although if they're as memorable as the ones you've given as examples, you could probably get away with it.
That all being said, I'd probably downvote such a question for being intentionally unhelpful and "clever". This site isn't a puzzle factory. We aren't here to solve Story-ID questions with as few clues as possible.
I'm not sure if this has ever been enshrined in meta law, but in practice it's what I've always observed.
If an ID question is uniquely answerable, then it can be left open.
If there are multiple answers which fit the description, then it's too broad.
(Interestingly enough, this is also the same criterion used for riddles on Puzzling SE, which actually have a lot in common with ID questions here: something is described via various details and answerers need to find its name. The difference is, as Valorum has already mentioned, that ID questions here are just meant to be answerable, not necessarily challenging.)
In the past we've allowed some very short ID questions to remain open precisely because they're uniquely answerable. Length shouldn't really be a consideration when deciding whether or not to VTC.
(I'm aware that other SE sites have found this policy impractical and vote to close questions based on 'quality' considerations as well as just answerability. But SFF hasn't had massive problems with ID question quality, and this is AFAIK the criterion most people use here for closability of ID questions.)
It really just depends on whether people can figure out the answers. These are not quiz questions, after all; the goal is to get people answers to their queries, not to prove anything about how clever or knowledgeable the answerers are. If people somebody can figure out the answer just from a single-sentence blurb, then that's all that is really needed. Obviously, well-known works are going to be easier to identify. Of the five on your list, I was able to get the first two without even a second thought.
Moreover, while there are guidelines for what makes a good story identification question, we ultimately cannot control how much information a questioner has about the work the are looking for. Some works are only very vaguely remembered, and some questions come to us second hand. It is not always possible to add more detail about when a book was read, or what the style of animation of a film was. The "best" questions are those that provide the most information, because they are the easiest to answer; and because they are will be the most useful to other people looking for the same work in the future. However, that does not mean that questions with less information are unanswerable; they are just harder.
Sometimes, a very minimal question will get a prompt answer, and the original asker will agree that the right work has been found. In cases like this, it might be advantageous to edit more information into the question and/or answer after it has been confirmed. However, by the most elementary indicator, the question has been successful: An answer was found. Thus, ipso facto, the question was not too broad or vague to be answerable.
