Very similar to an idea I have for a sub-optimizer. Sub-optimizer would seek a set of solutions in the top space, but not the tippy top. Problem is this tends to explode the amount of calculation and information to store but you can chop down the combinatorics dragon by segmenting the problem as much as possible into a series of smaller problems. My idea is first solve all the set/augment combos involving profane and solar etc. Then determine all the stacks of possible attributes, like HP, CON, Force spell power and individually solve for all the gear subsets that hit the top of each attribute, respectively, without any uncessary items in there that don't add anything. So you start not by solving it but by solving like 100+ mini problems. Then you create a tool to quickly as possible look at two or more subsets to see if they are incompatible, like they occupy same slot or they would result in more than two rings. Then use that to construct full sets seeking all the goals. By segmenting the problem, the combinatorics are reduced a bazillion fold compared to the full solution in one go. The result is not a solution but a big set of solutions that you could query without any priority issues. So you can do stuff like say I want DPS in the top percentiles (but not necessarily 100%) and I want good HP, good dodge, good MRR, good PRR, good AC, etc, but not as some trailing priority when its already too late.
