Another weakness not mentioned is document similarity search. You have one document and you want to find the N most similar in your collection to offer suggestions like Amazon does. You can do it with fts in pg, but it is way to slow to be practical. Custom fts engines are much better at that task.