Let's look at the different approaches for separating or combining this data and the pros/cons of each. Firstly, Esri provides standard models to get you started for Water, Wastewater, and Stormwater (links and descriptions can be found in this article), and for the purposes of this discussion let's assume you're starting with one of those models.
Your initial approach of putting them all in one network and separating them out by tier would imply that you aren't using one of the Esri models, which separate them out into different domains (Water lines, sewer lines, etc). If you wanted to do this you would need to make your own model (which is a lot of work!) where everything has a single domain and as Mike noted doesn't really align operationally with how you manage these systems. In this situation all of your layers would have a single set of rules, fields, and drop downs that all users would share. Users would be able to view/edit all the data in the network, so you wouldn't be able to restrict editing of water data to water users or vice versa.
The second approach often considered is to combine the different domains of models provided by Esri into a single utility network. This sounds good at first, but because everything exists in a single-network that means that all your viewers and editors from each department will need to be able to view/edit all the data in the network. So water people will be able to edit wastewater data and vice versa. The same can be said of the administrative experience. This does make it easier to manage/maintain, but once again it typically doesn't like up with the way that organizations want to administer their systems (in this case IT systems and databases, instead of pressure zones and sewershed areas).
Finally, the most common approach is to create separate feature datasets and utility networks for each domain. These allows the user/administration privileges for each domain of data to be managed independently of the other network domains. It is still possible to trace between these separate networks by using tools such as the "Add Starting Locations" gp tool to turn the results of one trace into the starting location of another trace.
