Some remarks on the stability of time-domain electromagnetic simulations