In our implementation, we keep this exclusive lock although the database management system also keeps its own locking on the level of tuples which is less restrictive, which would allow for more than one index update thread and certainly more concurrent searches.