Could you please write clear objective criteria that allow anyone to unmistakenly determine in advance whether his or her question will be deemed too broad?
I'm afraid it's not possible.
Stack Exchange network (and its model) started with StackOverflow, a Q/A site for software developers.
Software development is quite a special area, where you can take almost any kind of problem apart, isolate the problematic part, and present it to the world in the way where anyone can easily make the very same problem happen on their computer and solve it.
Most of the question requirements (general utility, reproducibility, and the one you're mentioning, the scope) have their roots in that StackOverflow model.
I should admit that this model does not translate very well to other fields of human endeavor, including linguistics. Language questions are not reusable the same way the software questions are, their utility cannot be defined as clearly, and the way people use language (to communicate in general) is not exactly the way they use software (to get stuff done)
However, we're still bound by the limitations of the network and the way the score system works.
We're not here for the karma farming, of course, at least I hope so, but the score model is an integral part of the network and is largely the reason it grew so successful.
I as a moderator try to stick to the rules as written, the way I see them, because it wasn't me who built this model and I don't feel confident enough with my abilities as a community builder to risk ruining the community we have built over the years, small as it may be, by introducing rules other than people more experienced in that area than me have developed.
I (I personally) feel sorry every time I have to close a question which is a good question by itself, but does not seem to me to fit into the site's model.
I feel even more sorry that I cannot come up with a clear rule set for the questions scope.
However, some decisions will always be left at moderator's discretion. The best I think I can do is be vocal about the reasons I performed some or another moderatorial task, and try and lead by example.