I think you have to be more specific about what you want to accomplish and define some rules that you want your divisions to follow. There are an infinite number of solutions if you considered multipart polygon divisions as acceptable for the solution. Division into ring buffer, horizontal, vertical, diagonal, cubic and squiggly line sections are all possible valid approaches to the problem, but if only single part polygon divisions are allowed some classes of polygons won't have any solution with a strict adherence to some of these approaches. So without some kind of rules or preferences that limit or rank the possible solution set the problem has no single right solution and that uncertainty only increases the complexity of validating any result as a solution to the problem.
