Slow terms aggregation speed on ~130M documents

I meant that whatever the user search for it should be fast after some calls.

Did you try that?