In Azure I can create
What are the practical consequences/differences? E.g in terms of routing, isolation...?
Well, it all depends on what you want to deploy, what your goal is. So I can only give some limited advice as I do not know your use case.
A reasonable approach is to create seperate VNETs for different workloads like dev/test/prod and have subnets in them for the different resources.
Then there is the Hub and Spoke architecture in which different VNETs containing workloads (spokes) can be accessed through a dedicated VNET (the hub) that controls the traffic flowing in and out. The hub is peered with the spokes.
The docs also list some considerations and azure virtual networking patterns.