The Asaro–Tiller–Grinfeld instability revisited