> What are we going to do, run direct fiber from every computer to every other computer directly?

No, you don't have to do that. A quantum network would be a web of point-to-point quantum links, with paths formed by routers choosing links. Same as a classical network.

To be a bit more concrete what an operating quantum network would look like is a bunch of routers using links to build up entanglement with their neighbors. When an endpoint wants to send a message across the network, a path from source to destination would be determined and entanglement across the links of that path would be consumed to move the message across the network [1][2]. The reason it's done this way, instead of directly sending the message, is that entanglement can be cross-checked before using it [3] and quantum networks really don't like dropping packets due to the no-cloning theorem.

> We typically don't call them networks until we start linking them all together with simple routing logic

Yeah I agree that it would be more accurate for this press release to say they made a quantum link.

> To me these are all just signs that the whole scheme is/was and will forever be mostly crankery.

Don't confuse difficulty with crankery. It'll be awhile before anyone reports an experimental realization of a true quantum network, because it'll be awhile because anyone can make a quantum router. The issue is that a quantum router is for all intents and purposes a fault tolerant quantum computer, and that is its own hard challenge being worked on separately. In particular, a quantum router needs to be able to store qubits reliably for non-trivial amounts of time, and to perform reliable operations on those qubits in order to cross-check stored entanglement.

[1]: https://en.wikipedia.org/wiki/Quantum_teleportation

[2]: https://en.wikipedia.org/wiki/Quantum_entanglement_swapping

[3]: https://en.wikipedia.org/wiki/Entanglement_distillation