I see the "sweeping change" argument put forward often, but I don't really understand how this can be the case.
There is no atomic deployment of a large distributed system - so even if you can check in related changes in different areas of a codebase, how do you release them?