Did lockdowns really save 3 million COVID-19 deaths, as Flaxman et al. claim?

Key points about the recent Nature paper by Flaxman and other Imperial College modellers