I work as a water network engineer. When I model water supply networks, I usually put sources as pressure nodes and sinks as demand nodes, because water simulation software solves for either the head or the flow at nodes. And I know the capability of the pumps at the source and the consumption at the customers. Flow in pipes is resolved with headloss-flow equations such as Hazen-Williams or Darcy-Weisbach.
In your example the sinks demand more than what the source can provide, all in terms of flow. The customers at both B and C will try to open their faucets as large as possible, to fulfil their 1 unit of demand; but assuming that the pipe path from a to B is identical to from a to C, the 1 unit of flow will split evenly after both B and C tried their best to maximise flow to their respective ends.
But because the constraint of 2 units total demand is not met, the simulation software will not resolve. Either the source should be changed to pressure node, which will give the pressure required to send 2 units of water out, or the customer demands should be reduced to match source capabilities. In the latter case, the purpose is to model the hydraulic grade line from source to sink.