Thanks. When it comes to "easily refactor" I think there's a distinction to be drawn between "hard but understandable and verifiable" versus "maybe-easy but with unknown surprises."

The difficulty of replacing or refactoring code should come from the "inherent complexity" of creating its replacement, not from the "incidental complexity" of discovering its boundaries and influence.