I have been in denial about this with the Bardi grammar for some time now, but the current revision work means I can’t ignore it.
There are several different approaches that can be taken to describing Bardi. One is to orient the grammar around the morphology. There’s a lot of morphology, so there is no shortage of forms to base a description around. The second is to base it around constructions, or your grammatical unit of choice. The third takes speech acts/functions as the organising principle. Of course, different aspects of grammars work better one way than another. But I have a whole hodgepodge of topics which don’t seem to fit together coherently.
One is possessive marking. Some nouns have possessive marking as a prefix; others only have a pronoun, but the order of element isn’t fixed, and possessives are also used for non-possessive meanings. Either a unitary set of concepts is split over multiple morphology-based chapters, or the possession marking is kept together but the contents of the section cover morphology, syntax and agreement patterns.
The sequence in which information is introduced is also problematic. It makes sense to put all the morphophonology and boundary stuff in one place, but it requires some explanation of clitics which hasn’t happened at that point.
This is not a new problem for grammar writers, of course.