Is there any proof of a lower death rate, really? In April, especially, so many people weren't getting tested, even if they were pretty unwell, let alone those with mild symptoms/asymptomatic. It's very likely in the first peak there were tens of thousands of new cases each day, but only a small proportion were being detected.
Whereas now we are detecting the majority, or at least a much larger proportion of cases.
Demographic factors are also possibly relevant- if most people getting ill right now are under 30, a much lower death rate makes sense.
There's also a pretty long lag on deaths-probably longer now, as we are catching cases earlier.
Spain's cases started to spike, early-mid July. Even into the first week of August, they had a 7 day rolling average daily death number under 10. Last week, that rolling average was over 30, and today it is 60 (admittedly there did seem to be an unusually large number of deaths on the 4th). Obviously, in real terms, these numbers are low, but the rise is concerning (to me).
But if we followed that pattern, we probably wouldn't see changes in death numbers until October at least.