Currently, 'Staytime' is increasing by 2 seconds each the predicted response.
Since prediction is not always requested every 2 seconds, 'Staytime' must increase in proportion to the accumulated queue length.
