Climate hazards are largely felt not through global mean temperature, but through extreme weather events, which are dangerous not only for their physical severity but for their rarity: by definition, they are very difficult to anticipate and prepare for. The same characteristic makes risk assessment a very hard statistical problem. Numerical simulations can be used to augment small sample sizes, but at great computational cost. Rare event algorithms offer a novel way to ``steer'' simulations towards the extremes to do targeted risk assessment at reduced cost, but this can be challenging when the events under study are transient in nature, such as passing rainstorms and heat extremes in Earth's midlatitude region. This paper presents the first successful application of a rare event algorithm to such transient extremes in an idealized model of Earth's atmospheric circulation, building on previously published results with a simpler toy model of spatial chaos. The core of the method is to select the right time to perturb the simulations, and the fact that the method generalizes is a promising sign that it can scale to even more complex, realistic models.