I once spent three thousand dollars of a client’s money on the wrong weight of lead came for a cathedral restoration because I was too embarrassed to admit I had misread the structural blueprints.
Most technical deployments are not built on logic; they are built on the desperate avoidance of re-doing work. But we call it optimization-even though the “optimized” system is actually just the one we stopped touching first-to keep the project timeline from hemorrhaging.
I am writing this while mourning the loss of I accidentally closed ten minutes ago, a digital massacre of research into medieval glass chemistry that mirrors the exact kind of professional frustration I see in IT infrastructure every single day.
In the world of Remote Desktop Services, this silence has a name: Andres.
The Ghost in the Server Room
Andres is currently sitting in a server room that smells faintly of ionized dust and ozone, looking at a spreadsheet he didn’t mean to create. He has just finished a grueling . The Windows Server 2022 environment is up. The applications are installed.
The RD Gateway is finally behaving with the firewall after three late nights of trial and error. Twelve department heads have been trained on the new remote workflow. Everything works. And that is exactly why Andres is going to keep his mouth shut about the fact that they bought the wrong licenses.
The retail difference between 50 User CALs and the necessary Device CALs for 14 shared workstations.
He realized it at of the . While auditing the final connection logs, he noticed that their warehouse floor operates on three shifts using only 14 shared workstations, yet the company purchased 50 User CALs.
If they had purchased Device CALs, they could have covered the entire floor for less than half the cost. He did the math privately. The difference is roughly $2,842. In a vacuum, $2,842 is a significant amount of money. In the context of a “finished” project, it is a ghost.
To raise the issue now would mean reopening the procurement ticket, admitting a fundamental misunderstanding of the licensing model, and potentially reconfiguring the Licensing Manager role. So, Andres closes the spreadsheet. He calls it “project completion.”
This is the sunk cost of a deployment. The more work we invest in a configuration, the more we treat the underlying assumptions as if they were carved in stone, rather than merely typed into a console.
“The hardest thing to rebuild isn’t the registry; it’s the consensus that we know what we’re doing.”
– Elena, Systems Architect
She was right. The reluctance to revisit a licensing choice isn’t just about the money; it’s about the psychological weight of the “Done” state. Here are the seven anchors that keep us tethered to expensive, incorrect licensing choices even when we know better.
1
The Momentum of the “Done” State
In my workshop, once the lead came is soldered at the joints, the window is a single, rigid unit. To change one piece of glass in the center requires melting down the perimeter, pulling the structural heart out, and risking a total collapse of the tension. IT managers view a “stable” deployment the same way. The license server is activated. The green checks are visible. The moment we achieve stability, we stop being architects and start being guardians.
2
The Social Tax of Correcting Upward
Andres knows that if he goes to the CFO now, he has to explain why the initial request was flawed. This isn’t just a technical correction; it’s a social risk. In most corporate structures, being “wrong but finished” is often rewarded more than being “right but delayed.” We treat the licensing purchase as a one-way valve. Once the capital is spent, the organization’s collective memory begins to seal around the event.
3
The Myth of Technical Debt as “Pragmatism”
We often frame our refusal to fix things as a pragmatic choice. “It’s already configured for User CALs,” Andres tells himself. “Changing to Device CALs would require me to rethink the Active Directory groups and the license server mode.” This is a lie. Reconfiguring the RD Licensing mode from “Per User” to “Per Device” takes less time than it takes to drink a cup of lukewarm coffee. The “debt” isn’t technical; it’s emotional.
4
The Complexity of the Choice
Microsoft’s licensing for Remote Desktop Services is notoriously opaque. You have to choose between User CALs (best for employees who use multiple devices) and Device CALs (best for shared workstations or shift patterns).
User CALs
Best for: One person, many devices. (Office staff)
Device CALs
Best for: One device, many people. (Warehouse)
When faced with complexity, the human brain defaults to the “safe” path. User CALs feel safer because everyone has a username. It’s only after the deployment is live that the waste becomes visible, like a crack in a window that only shows when the sun hits it at a specific four-o’clock angle.
5
The Illusion of the Perpetual License
Because these licenses are perpetual-meaning you own them forever-there is a sense that the money isn’t “wasted,” it’s just “invested.” We tell ourselves we might need those extra User CALs in three years. We treat the over-expenditure as a hedge against future growth, even when that growth is nowhere in the five-year plan. It is a way of soothing the ego by turning a mistake into a “strategic buffer.”
6
The Exhaustion of the 120-Day Grace Period
The RDS grace period is a double-edged sword. It gives you time to build, but it also creates a countdown that induces panic. By the time rolls around, the priority isn’t “is this the most cost-effective model?” but “will the users be locked out on Monday?” We buy what we can get our hands on quickly to stop the clock.
7
The Lack of a “Safety Net” Mentality
Most people assume that once a license key is issued, the transaction is over. In the stained glass world, once I cut the glass, it’s cut. But software doesn’t have to be that way. The fear of being “stuck” with a mistake is what makes Andres stay silent. If he knew there was a way to pivot without losing the budget, he might speak up.
I look at my stained glass windows now and I see every mistake. I see the three-millimeter bow in the lancet window. I see the lead lines that are slightly too thick because I was afraid of the glass breaking. To the parishioners, it looks like a miracle of light. To me, it looks like a series of choices I was too tired to revisit.
Andres will likely never tell his CFO about the $2,842. The server will run, the users will log in, and the wasted CALs will sit in the digital rafters like unpainted trim in a house. But it doesn’t have to be that way for the next project.
If we want to build infrastructures that actually last, we have to be willing to reopen the crate. We have to be willing to admit that the “pragmatism” of the present is often just the exhaustion of the past.
I’ve finally reopened those forty-three tabs. It took me an hour of searching my history, but the thread is back in my hands. It was annoying, and I felt like a fool for clicking the “X” in the first place, but the work will be better for it.
Don’t let your deployment be defined by a mistake you made in the eleventh hour of a grace period. Reopen the question. Change the CALs.