I’m not sure why a child-comment got killed: >>> Not true. Deaths only lag cases by 5-10 days. For example, in Italy daily new cases peaked on March 21st and daily deaths peaked on March 27th.

This is correct. People are expecting a step function of deaths exactly 4 weeks from the peak of cases? This is not statistical. Peak of deaths lag peak of cases closer to 1-5 days. If case growth is a matter of true rise of infections we would see modest increase of deaths about 2-3 weeks ago (cases started growing 4 weeks ago). We’ve only seen deaths drop on a exponential decay trend.

Isn't it plausible this could be affected by other the changes in infection demographics, testing capability, and treatment options we've seen since then?

Yes, it’s related to all those things including the never-talked-about problem that the lockdown is ineffective at suppressing the inevitable spread of corona

Cases started increasing above norm 4 weeks ago.

Not true. Deaths only lag cases by 5-10 days. For example, in Italy daily new cases peaked on March 21st and daily deaths peaked on March 27th.

This assumes the death rate is independent of, say, health care system capacity. Didn't they have an issue with that in Italy?

Doesn't that depend a great deal on testing capability? If testing is quick and widely available, then I'd expect the lag to be larger as people get diagnosed earlier.