It isn't helpful to describe network solids as molecules, so we probably shouldn't
The problem here is that there in't a strict definition of molecule in chemistry. So we have to decide whether particular uses of the term are helpful or not.
We could argue that any entity where all the bonds are covalent is a "molecule". This would include diamond and polymers. But this is not helpful in practice. polymer scientists don't, for example, find it helpful to describe the components of a polymer as a molecule (they might use the more specific term macro-molecule). Most non-crosslinked polymers don't consist of single "molecules" but a mixture of many entities with different chain lengths. Talking about "molecules" isn't much use in describing or characterising the overall polymer. Many properties depend on the statistical distribution of chain length and thinking individually about each specific chain is unhelpful and irrelevant.
For many substances it is helpful to talk about the molecules making up the substance. Usually this is because the molecules that make up the substance are well defined. We can exactly describe the atomic makeup of each molecule making up the substance. Every molecule of aspirin is $$\ce{C9H8O4}$$ and no matter how much we have that describes the substance in a useful way. Where a substance is made up from more than one molecule, this is still true.
The problem with network solids is that there is no unique molecule. Every single crystal contains a different number of atoms; every diamond is different. So the idea of being able to describe the common feature of the substance as being the properties of the components (molecules) making it up goes out the window. It is better to call them network solids and think about the properties as a product of how the atomic components are connected. In mineralogy, for example, there are a lot of network solids. Silicate minerals, for example, can contain silica 3D networks (like diamond, in fact), flat silica sheets, linear silica chains and small silica units. Describing these components as "molecules" adds nothing useful; recognising the structures as components of a network is clearer and more useful.
In conclusion describing the components of network solids as "molecules" doesn't help so it isn't a useful use of the term.