The behavior of our components is largely self-contained, if implemented well. What they expose to the outside world are preferences that make subtle adjustments to their behavior. This could be something that's internally-focused—such as the type of model that's used, or the preferred algorithm. It could be something that's user-facing, such as enabling components, or setting display modes. It's these preferences that help us scale our components to work in a variety of contexts.
Once our software reaches a certain critical mass, not all features will be relevant to all users. The simple ability to toggle components between an enabled/disabled state is a powerful tool. Both for us, as a software vendor, and for our customers. For example, we know that some features are required by certain user roles in our software, but they're not the common case. To better optimize for the common user, we may choose to disable certain advanced features...