Some thoughts:
- It's still too early to measure the duration and frequency of long-term complications for a new virus, but early evidence suggests they could be important.
https://www.theatlantic.com/health/archive/2020/06/covid-19-coronavirus-longterm-symptoms-months/612679/
https://www.vox.com/2020/5/8/21251899/coronavirus-long-term-effects-symptoms https://twitter.com/NateSilver538/status/1272885634202570755

We can look at SARS or MERS and see how frequently these patients experienced long-term morbidity, but these viruses are also different. Prospective studies are being set up for COVID, although I imagine we will need more of them. https://www.hkmj.org/abstracts/v15n6s8/21.htm

- IFR isn't a very useful summary statistic, as it is an average across different age/risk groups, and the age/risk profile varies from place to place. I haven't seen evidence that it is declining, but that could occur if care improves or we more effectively protect the elderly.

I would tend to advocate for an approach that stratified by age to separate out these questions. Is the profile of those infected changing? And, once infected, are people better off?

- Surveys suggest that most places still have seroprevalence below 10%, unless they have had large outbreaks. There may be cross-immunity with other coronaviruses that wouldn't be reflected in a SARS-CoV-2-specific antibody test, but we have to wonder about the magnitude.

Certainly prior circulation of other coronaviruses did not protect places like NYC or Italy from experiencing large outbreaks. Cases are rising in Saudi Arabia, where MERS has been most prevalent. So cross-immunity could help certain pockets, but it is not a strategy.

- Regarding heterogeneity and herd immunity, there are some nice replies by @joel_c_miller. I made this thread in May. My feeling is still similar. I haven't seen convincing evidence that heterogeneity could dramatically lower the threshold. https://twitter.com/nataliexdean/status/1259248274625761282

It relies on the assumption that the people most likely to get infected themselves are in turn the ones most likely to infect others. But is that true? Also, if the highly connected nodes are essential workers, and they get sick, won't they be replaced? The dynamics are complex.

Again, we can study the implications, but it is not a strategy. https://twitter.com/joel_c_miller/status/1272889883447775238?s=20