In areas with relatively good job markets, such as CS/Stats/Economics/Finance, Postdocs are often quite prestigious and are marketed by top schools as a way to attract promising young scholars and give them time to research without the pressures of teaching/service. In areas with very bad job markets, such as the humanities, they are also seen as prestigious because any academic position after the Ph.D. is extremely difficult to come by.
For the lab sciences, my understanding is that a postdoc actually is more similiar to doing a doctorate in another domain. Experiments are too expensive and time consuming to let a Ph.D. student design and develop a line of research on their own. Instead, that's what the postdocs do. Also, it is not an option. You must do a postdoc (often a long one, and often multiple) in order to have the experience needed for a TT job.
For mathematics and much of the social sciences postdocs are awkardly in-between all of these. They are increasingly common and almost required and increasingly long. Yet, there's no good place for them in the current system, where they are technically fully-credentially academics but treated like overgrown students.