Thanks, looking forward to reading the blog post!
I have another question in the same vein. Instead of applying a binary decision to remove the expensive rule variations, have you considered/tested using a weighting method like inverse cost proportionate (with reasonable boundaries)? Expensive...