ST_Subdivide divides complex geometries into smaller rectangular chunks of relatively similar vertex counts, which are then much better to use in spatial indexing, predicate, and related functions. The algorithm works by recursively cutting geometries along one dimension then the other.

This is not strictly about exposing GEOS internals to Python, so it might not be a good fit here. But given the recent work with Cython and strong dependence on internals here, it might be a worthwhile experiment (plus, I need this for one of my projects).

Unless there was opposition, I am hoping to try and implement this within Cython here.