As briefly as possible: Containment only has real utility if we can develop a game-changer that drastically alters COVID-19 outcomes, and that we don't have to wait so long for it that lockdown-induced mortality eclipses the time-adjusted benefit. It's a risky strategy that requires waiting for an uncertain, temporally-unbounded event, which in turn means that the "indefinite postponement" strategy incurs unbounded cost. We simply don't know enough to make definitive predictions about upper bounds on time-to-discover-incredible-treatment.
Whereas in a no-lockdown, natural herd immunity strategy, we are experiencing a fairly well-bounded spike in short-medium term mortality. I've been using Ferguson's 2.2 million figure for the US which is based off 82% population infection with an IFR of 0.9%, with the population equally susceptible at the physical level (behavioral differences are modelled, I believe). That's an excellent upper bound, it's certainly not the mortality we should _expect_. So we have a much tighter certainty, we will experience somewhere between 200,000-2.2 million deaths due to COVID-19.
When weighed against the alternative, which again is to incur unbounded risk from lockdown-associated mortality, the correct path becomes quite clear in my opinion. A very marginal increase in all-cause mortality in the whole population could very easily offset the spike in COVID-19 mortality which is constrained to fairly limited subsets of the population overall.