I think it has to be something else, maybe one of those 3 things I hypothesized.

The reason I say it looks like herd immunity is that what GreenevilleVT describes 6 weeks to peak and 6 weeks to minimal new cases describes what we expect. That is the derivative of a logistic curve. The reason total cases is modeled with a logistic curve is that at some point (inflection point and peak of derivative) the odds of infecting someone start going down and at another point there's no one new to infect, so there must be an asymptote (a maximum value in total cases) somewhere. With Herd Immunity, the asymptote is some majority of the population (what are they saying now? 70%?) where the virus dies out because no contagious person is near a person who can be infected. Of course, the odds of transmission can be changed by changing behavior not just by the number of people who have already had it.

For some reason, if we stay near "minimal new cases" we are at that asymptote. Is the peak of the derivative a local or global maxima? Why are we there earlier than we anticipated?