[m0llusk]

This misses the problem of scale. Often the only way to patch an existing system is to try small and localized changes and then propagate them slowly. Even correct changes may require unexpected related details to be in order for a successful result. Going slowly allows for analysis of related factors and anticipating range of variation.

[Kamq]

This reminds me of Eisenhower's old quote to his advisors: “Let's make our mistakes slowly"

Which, ironically enough, if you make changes slowly, and have a good way to back them out when they go wrong, you can change much faster. Mostly because you don't have people screaming about a change being the end of the world so hard that you need to go to war in order to change the smallest thing.