Discussed with Melanie Courtot the possibility of splitting an abox for more scalable reasoning.
The splitting algorithm requires some finesse to keep completeness, but in some cases a good tradeoff can be obtained, depending on the input. A framework to allow for this should be useful.