Nuclear physicist here, so I can't hang when it comes to crunchy immunological details, but here's the gist as I understand it:
The vaccines use a designed and stabilized version of the spike protein, which leads to antibodies that bind more strongly and thus more effectively. People aren't likely to be reinfected by the same strain, but mutations can reconfigure the spike protein slightly, making it bind more strongly to cells, and less strongly to those antibodies.
"Naturally" derived antibodies may be as good as the ones from a vaccine, but that's a gamble. In aggregate, people who have been previously infected (with a strain other than delta) are more likely to catch/have poor outcomes from delta than people who are fully vaccinated, by a factor of 2-3. This is a fairly recent result, and I can't find the reference, sadly.
The absolute risk is low, but the relative risk is huge. Also, when you look at the total risk budget, again, vaccinated immunity just swamps natural immunity. It's only the naturalistic fallacy to think otherwise.