This dynamic can be seen by plotting the CDC COVID-19 US population vaccination counts as proportions of the US population size alongside the NYT COVID-19 infection 7-day trailing average cases counts for the US population using a dual y-axis for the two public use data sets.
I found a bug in the reported CDC vaccination data for early June for the raw case counts. Specifically, the case counts reported were -24,000 cases. Hence, I switched to the moving average for the case counts. The lesson here was that I thought I had not correctly coded the axes and spent some time trying to fix it. When in fact the issue was an error in the data reported. Hence, check your data source :)