Herd immunity for a disease with an R naught (R0) value of 2.5 kicks in at 60%. The formula for calculating herd immunity is 1-1/R0. The R naught of Covid-19 is between 2 and 3, so I use the value of 2.5.
For the US, that means around 198 million people would have to be immune for herd immunity to set in. We don't know how long immunity from the illness lasts; some studies suggest that it is a few months only. In any case, the number of people who have recovered together with those who have been vaccinated is not even close to 198 million. Any decreases in cases and deaths at this point are still due to measures to slow spread, such as masks and social distancing.
Nearly 30M people in US have confirmed-by-test COVID infections. It’s common to estimate at least 6 undiagnosed cases due to asymptomatic cases, people choosing to wait-and-see rather than get tested, etc — for an estimate of 180M people who have already been exposed. That’s quite a lot of poeple. Herd immunity isn’t light a light switch that is on or off — more like a dampening knob that is steadily being turned toward 0.
You make the fatal assumption that all cases are counted. Recent literature states we are between 6.5 and 8!x under the true cases. If this is accurate thst means anywhere between 140- 160 M cases and add vaccination and 190 M is easily cleared. The graphs and trends do look like a herd has developed as mitigation has been in place over a year snd we still had the strep clumps 2 -3 months ago
As for no idea on the length of immunity certainly if immunity was only 2-3 months the literature would be rife with reinfect in data. It is scant. The chances is reinfection appear quite low.
In this case the observation days favor that a herd is developing and this is not related to mitigation factors.