Like "token sorceries" or "30 sagas" or "those two characters" or "there could be more split card mechanics besides fuse".
I have no idea how people finish things because I tend to get about half or a third of the way and then lose the spark that made me make that first part; or realize I don't quite know what I want out of it, or that some core part of my inception is flawed.
I usually write a cool card and turn it into a mechanic that gets spread around.
I start with mostly the commons, but I always jump around to uncommons/rares/mythics because it just seems the most sensible.
Amount of worldbuilding depends; frequently I tie the two parts together: I want this card or I want this event so I engineer the other to make that feel right.
Of course; since I have nothing completed to show for it, perhaps my words are but that of a fool.
This is a tough question as different approaches work for different people imho
Over at the old forum, we had this useful guied I'd suggest taking a look at - even if some of the stuff mentioned didn't really work out for me personally in the end, it really helped trying nethertheless.