In reply to nikoid:
> I was thinking along these lines but the undiagnosed cases (as I see it) have no bearing on the closed case rate. That's why the closed rate case seems to be a more accurate way of looking at death rates. Still scratching my head though.....
This would be true if the under-testing was still random. As is, severe cases will be getting identified because they turn up at hospital, where mild cases don't get diagnosed. The outcomes for the hospital cases are worse than the non-hospital cases. That drags the average down. Or to put it another way, pretty much all the deaths are getting counted, and a lot of recoveries could be getting missed.
I think it's probably a combination of that and, as MeMeMe says, the different recovery times. While the number of cases is growing quickly (probably still exponentially) the statistics will be dominated by recent cases just because there are more of them. And the fatalities get resolved quicker than the recoveries, so drag the numbers down. As the infection rate peaks and starts to drop off again, we'll see the number of old cases exceed the number of new cases, and the mortality rate should come back down (if nothing else has a bigger effect).