JPA is basically an abstraction using ORM methods. If you map various model classes to a database, JPA can: a) generate the appropriate SQL query / update, b) convert the resulting groups to model classes. JPA also includes caching and abstract transaction processing.
In the end, it does nothing magical - it all ends up passing through your JDBC driver, becoming raw SQL and returning results for JDBC, etc. It just allows you to hide a lot of code and just work with your model classes like regular Java objects (POJOs), where setting a property starts UPDATE and getting a property starts SELECT (caching everything and the organization in a transaction provides much better performance than you could get through simple one-to-one implementation.
So your second question makes no real sense - if 1000 queries come out, it is mainly a DATABASE database server that should scale and process this, not JPA. (Admittedly, he should deal with sending them and then sort them into java objects)