The current implementation indexes documents in single fields (title, doccontent, etc.) and analyzes them with a generic (so-and-so performant) cross-language analysis chain. This results in poor index and query performance compared to a specialized, language-specific analysis.
There are several solutions that can be implemented and, from the discussion, we have concluded that using the multiple-fields solution (title_en, title_fr,..., doccontent_en, doccontent_fr,...,etc.), coupled with some query post processing (for easier query writing) and configuration auto-generation (for easier setup), is the one to go with.