Writing a design system that survives its second year
Year one is components. Year two is everything the components could not predict, which is where systems die.
The first year of a design system is easy: name the tokens, build the buttons, feel good about the Figma file. The second year is a queue of requests for a variant that does not exist.
Systems survive that year by making the escape hatch official. An explicitly supported way to break out is cheaper than twenty copies of your button component with one property changed.
Count the forks, not the adopters
Adoption is easy to measure and tells you nothing. The number of components teams copied and edited is the number that predicts next year.