Cheap/low end doesn't have to mean sucky and dangerous. Look at what happened in IT. When I started, many IT guys were strong generalists, capable at a wide variety of tasks. It was not uncommon to see a single team handle design and architecture, development, testing, requirements gathering, deployment,and support. However those guys were fairly expensive and managers figured that it would be better to compartmentalize the work and hand it either to specialists for improved quality (or at least repeatable mediocrity), or to lower paid workers to handle the simpler tasks like 1st line support.
Over time the different components of a datacenter became so complex that you had to specialize. In large companies, Networking and SAN storage management require their own specialists and departments.
But, you still need high end people in those departments. Just because they are specialists, that doesn't mean they can be mediocre. The problems I have seen over time is that when one component has a problem(Storage, Networking, Server, Application), there is a lot of finger pointing. I have seen storage performance issues where the storage team claimed there was no problem with the disks and meanwhile the FA ports were running at over 90% utilization, which is bad if you were expecting good response times. I have seen countless networking problems and the networking group would claim there was no problem with the network, meanwhile it turns out that they made firewall changes that were blocking required ports. DBAs often lthink there is a disk performance issue when the real problem is a bad query and now it's doing a full table scan on a 4TB database, that never ran in 5 seconds!
All of these problems should have been noticed within minutes, but mediocrity makes it possible for these problems to last for months because someone doesn't see the obvious problem and the other teams have no visibility into that mediocre team's environment. To compensate, there need to be high end players with visibility to all of the environments so that they can point out what someone is missing. Paying for people like that seems expensive, but it's not nearly as expensive as multiple teams spending months on a database performance problem that only one team can solve, but they keep overlooking the obvious problem and nobody can call them on it.