Get the policy table with the solution.

This is cleaner, but means we are building the same inference engine many times, it might be time to create some caching logic for the engine to make the solution quicker for larger project

No file, test, or entry point is named. Start by locating where the inference engine builds the policy-table solution and how often that path runs for larger projects. Determine the cache's scope and invalidation behavior with the maintainers, then verify that repeated inference reuses results and improves performance without changing the solution.

Written by the indexing model from the issue text.