We explicitly have one ranking for each index. This is done for performance reasons. If you've not seen it, this page (https://www.algolia.com/doc/guides/managing-results/refine-r...) in our docs gives some guidance.

Have you looked at using virtual replicas (https://www.algolia.com/doc/guides/managing-results/refine-r...) at all? They don't increase the number of records as they use the same underlying data.