This is also the same format done by Radiolab. I don't know if this is a well-documented structure to be honest.
I derived it myself after becoming lethargically bored listening to these types of podcasts. I realized they all had essentially this structure, sometimes with a few steps removed but rarely, almost never, out of this sequence.
I would like to claim, although a lot of data would need to be collected, that these steps have narrow time ranges with normal distributions and fairly tight bounds. As an extreme example, they don't spend say, 53 minutes on machine sounds and then cram the rest in the remaining time. I think there's more or less a clockwork to them.
