When I make a campaign world I like to start with the cosmology. I also like to commit to the cosmology as much as possible. I think this can produce interesting setting elements that deviate from our own world in ways that make the place more compelling to players.
Sometimes this also means not adhering to modern assumptions about how the world works. That can be difficult because we have a tendency to cleve to the real world when possible, which makes a certain amount of sense. After all if you deviate too much your players will feel adrift in a confusing sea of things that have no connection to their own experiences. So you do have to be careful. But I don't think we should be afraid to experiment with big changes and to commit to the implications of those changes.
For example, in a world where you assume the existence of gods and those gods not only shape the geography but control things like the weather, are realistic weather patterns necessary? They may be depending on how much of an active role you see the gods having in managing climate and weather but they do not have to. This is a space where committing to a cosmological assumption can produce some interesting things. If it really matters for the harvest that the people keep Lord God Lomkha of the Mountain pleased, and the rains haven't fallen in some time, a quest to discern the cause his displeasure is certainly a possibility. It means if the players encounter a big storm that didn't happen because of the rotational movement of warm and cold air in the atmosphere, it happened because of the gods. Maybe your god of storms sent it as retribution, maybe it was an aimless fit of rage, or perhaps it means several gods are in conflict with each other.
Cosmology can liberate you by making things possible you had never before considered. But it is also a restraining force. When we designed Sertorius, assumptions about the nature of magic, souls and creation gave us fuel for some key concepts. Grims, which are strange supernatural features of the landscape were a direct result of the cosmology. So too was the existence of the halfling mercantile Republic of Shahr (being favored by the Sea Goddess gave them a key competitive advantage). These were the liberating aspects of cosmology. The flip side was we had to adhere to the constraints of the cosmology when we made things like monsters. We couldn't just make whatever we wanted, each creature needed to fit within the setting assumptions. That proved challenging, but it also guided us into some interesting places.
When we made Sertorius, we committed to the idea of cosmology to a degree, but I wish we had committed more (particularly around things like weather and geography). I think we did a good job of establishing a cosmology and keeping an eye on it, I just see wrinkles in hindsight I'd have smoothed over. When we started working on Wandering Heroes of Ogre Gate, I made cosmology a priority.
The setting flows from key assumptions about the nature of reality in Sertorius. The setting of Qi Xien is an outgrowth of events in Gamandria (both directly and indirectly). There is a Heavenly Bureaucracy in Qi Xien inspired by Chinese myth and at its head is The Enlightened Goddess, who is the being described in Sertorius as Aetia. These are identical entities and the reason she created Qi Xien was to establish a place more perfect than Gamandria (which was nearly destroyed by her petty gods). So most of the setting elements are a product of her desire to make something that didn't succumb to the troubles of her previous creation. Of course, conflict making for a better setting, something slipped through the cracks and introduced a bit of chaos, and this serves as the foundation for the game. Just like we had to explain the existence of Magic and Sertori in Sertorius, in Wandering Heroes of Ogre Gate we had to explain the existence of Qi and gravity defying Martial Heroes.