A mature product has history. Customers have workflows built around it, teams have operating habits, and the product has accumulated assumptions that may not be obvious until you try to change them.
That makes evolution different from starting from scratch. A cleaner new model can still be the wrong move if the transition creates too much disruption.
The hard part is sequencing
I find it useful to separate the destination from the migration path. You can be clear about where the product should go while still introducing the change gradually.
That might mean supporting old and new workflows for a while, moving one behavior at a time, or testing a new model with the customers who benefit from it most before making it the default.
The important thing is to treat current behavior as evidence, not baggage. Existing customers are showing you what is valuable today, even if the product needs to change for tomorrow.
Progress is not replacing the old thing as quickly as possible. It is getting to a better product without making people regret coming with you.