The Hadoop RMI implementation enables distributed search on SOLR. The advantage over the traditional SOLR distributed search is the ability to easily use alternative queries such as span queries or customized queries, and use custom analyzers. Hadoop RMI implements dynamic classloading without the need to jar the classes and make them available over HTTP. LUCENE-1336 has example code that illustrates how reopening an index returns a remote references to the new searcher for that index.