Phase transitions in large scale systems resulting from changes in the average behavior of individual elements are a real thing. An infection is an unstable dynamic system which tends to different states depending on whether the mean R is greater or less than one.
Also, remember that the effective rate of reproduction can easily vary by huge amounts between different parts of society, for example if some age groups are almost completely vaccinated and others are almost completely unvaccinated.
Presumably, Interesting things happen around 1.0, where the instability can push you either way. Hence heard immunity is almost a binary state.
R > 1: Infections rising R < 1: Infections declining
If we remove all restrictions and remain R < 1 due to vaccination then we have by definition reached herd immunity (key word is herd, not individual).
Take the US...
Certain sub-populations could very well have herd immunity within that group. A group like meat-packing plant employees and their families. They've all had COVID or been vaccinated.
But, within the group of "white collar workers who can afford to remain socially distant", we could be a long way from herd immunity - this group has remained distant (low number of past infections) and hasn't been eligible for vaccination (<65 years old, mostly healthy).
Yes, at a macro level, herd immunity is mostly binary. But, over a population of 300,000,000, there's a lot of room for sub-populations with differing levels of immunity.
Assume for convenience that Covid cases double each week in schools and halve each week in adults, there's no spread from one group to the other, and almost all cases are in adults initially. Initially we measure R and it's firmly below one, because almost all cases are in adults and those are declining rapidly. After one week there are twice as many cases in children and half as many in adults, after two there are four times and a quarter, eight times and one-eighth... and pretty soon the number of remaining cases in adults are so few any further reduction can't possibly compensate for the exponential increase in schools, and overall cases are growing exponentially and fast. Our declaration of victory was dangerously premature.
The exact numbers don't even matter; so long as the reproduction rate is firmly below 1 in one group and firmly above 1 in the other, the group where cases are shrinking will make up an exponentially decreasing proportion of total cases and the overall rate of spread will quickly become determined by the group where cases are still growing. (The assumption of no spread from one group to the other doesn't matter either; it just makes the explanation slightly simpler.)