Usecase is the following:
- There is a save() containing multiple structured documents (like PDF, ODT and ect);
- When saving into index, JCR creates a thread pool executor, that extracts text from documents in parallel.
But, when replicated volatile index was introduced, features stopped working. Root of it is a serialization of Lucene Documents. The last one retrieves the content of Lucene Documents invoking text extraction. This cycle performed before Lucene Document is being placed into index, before thread pool executor used. So there is no performance increase made by thread pool.