So what about the other two? First we need to know about the difference between entangled and unentangled quantum networks. QKD can run either with or without entanglement.
I'm lumping several different things together in the category of sensor networks. Examples include high-precision clock synchronization, improved resolution in arrays of telescopes, and the like.
So the sensor networks are attractive but difficult, especially when doing the engineering of the classical interface and control and worrying about noise.
Especially, we need to understand that there are similarities and differences between system interconnects or data center networks and wide-area networks. https://www.osti.gov/biblio/1900586
It's critical to know that scaling up quantum computers requires entanglement between quantum processors. If we want to use two small quantum computers to solve one larger problem, we MUST be able to create inter-node entanglement.
Beyond that, what about wide-area entangling networks? One of my favorite ideas of the last two decades is blind quantum computation, by Broadbent, Kashefi and Fitzsimons. https://arxiv.org/abs/0807.4154