One of my growing interests is in split syntax.

Though much of linguistic theory has tried to reduce the "rules" of a language to a small uniform set, it more accurately appears that different parts of a grammar may operate according to different generalizations.

Split ergativity has long been acknowledged.  But ergativity-nominativity is just one phenomenon that can be "split" across different parts of grammar.  For a few more examples:

In Panare, VSO order appears to be the norm in one tense-aspect; but OVS is arguably the norm in a distinct tense-aspect (Payne 1994a).

Also in Panare, NP syntax is highly non-configurational, while main-clause structure could be argued to be really quite configurational (Payne 1993c).

In Maasai, some of the nominal vocabulary operates on a lexicalized semantic gender system; while other nominal vocabulary is sensitive to a more "on line" cognitive-pragmatic gender system (Payne to appear d).

What does this mean for theory?  In part, I believe it means that explanatorily adequate theories of syntax must be historically-informed, acknowledging that change moves gradually through the grammar of a language.