Normal, unfortunately. See my post I made 2 days ago sharing graphical results of stress testing I did on assistants/thread model. Run time varied between 10-40 seconds… Not at all viable for a production app.