Back to the monolithic scenario
In the previous scenario with the spelling correction, it is clear that we failed with respect to the separation of concerns. We didn't have any modularization at all, at least from a deployment point of view. The system appears to have the undesirable features of low cohesion and high coupling.
If we had a set of separate deployment modules instead, our spelling correction would most likely have affected only a single module. It would have been more apparent that deploying the change was safe.
How this should be accomplished in practice varies, of course. In this particular example, the spelling corrections probably belong to a frontend web component. At the very least, this frontend component can be deployed separately from the backend components and have their own life cycle.
In the real world though, we might not be lucky enough to always be able to influence the different technologies used by the organization where we work. The frontend might, for instance...
