If it wasn't about money, surely they could have a second index for old code? And then asynchronously query both, filling the fast results in first.

Or maybe have a keyword for searching old code, or search if current results seem bad, etc.