You summed it all up. A redundant network requires more hardware to build and more qualified staff to maintain. Obviously, a simpler network is easier to maintain.
By scalability, there isn't much difference, but a very large network that has hundreds of SPOF would be a nightmare. At the same time, a large, _entirely_ redundant network would also be quite a feat, so practically you make a network redundant where it really matters (core, distribution) and leave the access layer and less important areas (with limited failure impact) mostly simple.