Since the pooling logic and keeping track of which connection is in use etc... happens within the wrapped proxy connection that you get from the datasource, I can see why all that logic would be skipped if you use the underlying connection for direct connection management. I would instead suggest changing that code to:

@Test 
  public void testRelease() throws SQLException { 
      OracleConnection c = null; 
      OracleCallableStatement s = null; 
      final Connection wrappedConnection = dataSource.getConnection();
      try { 
        c = wrappedConnection.unwrap(OracleConnection.class); 
        s = (OracleCallableStatement) c.prepareCall("{ ? = pak.func(?) }"); 
        s.registerOutParameter(1, Types.INTEGER); 
        s.registerOutParameter(2, Types.VARCHAR); 
        System.out.println(s.execute()); 
      } catch (SQLException e) { 
        e.printStackTrace(); 
      } finally { 
        if (s != null) { 
            try {
              s.close(); 
            } catch (Exception ignore)  {// ignore or log}
        } 
        // close the original wrapped connection
        wrappedConnection.close();
      } 
}