This is frequently done in large corporations too - divisions which don't directly serve customers, but instead provide enabling services to other divisions, make "internal accounting" charges against the budgets of the divisions they serve.
This is useful, in that it can let a support service (like, say, IT) prove its use in the regular way that businesses do (profitability). It also keeps other divisions from treating internal resources that are costly to maintain and provide as "free" - if they can find the same service outside of the organization for cheaper (i.e. in this case through Google), they have an incentive to use it.
The issue here seems to be that the NTIS is unprofitable, and it's not actually essential to the functioning of any other parts of the federal government.
tl;dr: Large organizations usually work in practice as a cluster of smaller organizations participating in a free market.
Ran in to this a couple of times over the last few years. The problem is that the internal group doesn't usually really operate as if they are in the free market, they just get to operate like a monopoly. I had to work with different orgs who were mandated to use the internal IT dept for all projects, but the IT dept had no incentive to actually do any work.
When you needed a server - that will take 3 weeks to set up. Really? It's going to take you 3 weeks to provision a new VMWare instance (cause 90% of the servers provided were VMWare instances). Really? OK. Hrm... There's a typo in the hostname - can we get that fixed? "No, that will require too much work updating our network to accommodate that."
Hrm... OK... Can we get an account for the outside contractor (me) to get on the VPN? "That may be 2 weeks - the person that does that is leaving for vacation on Thursday." Umm... It's Monday now. "There's a lot of stuff to get done before they leave."
I get that there's some benefits to having this centralized - security updates, monitoring, etc - but the impact of waiting weeks for things that should take a few minutes with any company in the real outside world is real (although hard to measure). When the entirety of the org is constrained this way, there's no comparisons to be made to make your case that it shouldn't take this long; many projects should be measured in weeks, not fiscal years.
FWIW, the times I've run in to this in the past... 15 years or so, it's usually academia. My understanding is that this is its own particular brand of IT hell.
The company I'm working for suffers from the same problem, but it's compounded by the fact that our IT is now almost 100% offshore. This means on top of having to wait for new VMs, we also have to fight a language barrier to explain what we need and why.
As a developer who has to wait weeks for instances that I could spin up myself in a day or less, and instead has to wait for literal weeks, I really don't see how this cuts costs. Perhaps, as you mentioned, it makes monitoring the systems easy, but for internal instances that have no internet access, how valuable is this?
This is useful, in that it can let a support service (like, say, IT) prove its use in the regular way that businesses do (profitability). It also keeps other divisions from treating internal resources that are costly to maintain and provide as "free" - if they can find the same service outside of the organization for cheaper (i.e. in this case through Google), they have an incentive to use it.
The issue here seems to be that the NTIS is unprofitable, and it's not actually essential to the functioning of any other parts of the federal government.
tl;dr: Large organizations usually work in practice as a cluster of smaller organizations participating in a free market.