What Chen actually said
Chen's point is that when a system becomes CPU-starved, every process suffers. That includes the taskbar and Task Manager, even though they are often the tools you reach for to stop whatever is causing the trouble. His post ran on The Old New Thing on October 27, and Neowin picked it up on November 16.
His core argument is about waiting. Higher scheduling priority cannot bypass locks or blocked shared resources. A high-priority thread gets to run sooner, but only if it is ready to run. If it needs something another thread holds, it sits and waits.
Chen's analogy is a priority checkout pass at a supermarket. The pass gets you to the front of the line. If the person at the register is paying for a huge cart of groceries, you still wait.
Priority inheritance doesn't remove the wait
Windows code can guard against priority inversion by passing a waiting thread's priority to the thread that owns the resource. Chen's point is that this helps the owner get CPU time, but the waiter is still stuck until the owner finishes. The owner might be doing something slow, such as slow network file access.
So the fix doesn't make the lock free any sooner. It only stops the owner from being starved of CPU while it works.
The UI system is its own bottleneck
Chen also singles out UI code. Sooner or later it has to call into the UI system, where it competes with all the other UI threads in the system. The taskbar and Task Manager are UI programs, so they hit this wall whatever their priority is.
This is a general design limit. Chen doesn't name a Windows build, version or specific bug, and he doesn't announce a fix or a setting.
"Hard to fix a runaway system from within"
Chen's broader point is that it is hard to fix a runaway system from within that same system, because the tools you would use may already be compromised. If those tools share critical resources with whatever is out of control, they can stall along with it.
Windows does already favor one thing. It gives preferential treatment to the program in focus, so getting focus onto the taskbar or Task Manager is about as well as you can do. That treatment only goes so far if those preferred threads are blocked on things controlled by non-preferred threads.
Foreground favoritism is already built in
Microsoft Learn's Quality of Service documentation fits this picture. It says scheduling priority is still the main metric Windows uses to pick the next thread. QoS can also influence core selection and power management. Windowed apps in focus get High QoS, visible apps not in focus get Medium, and minimized or fully occluded windows get Low.
QoS is about performance and power efficiency. It doesn't guarantee that the taskbar or Task Manager stays responsive, and it doesn't unblock a thread waiting on a lock.
Evidence from the comments
The comment thread under Chen's post is a useful reality check. One reader said big builds make the taskbar nearly unusable, and that raising Task Manager's priority helped only slightly. By their own account, that reader also lowered the priority of the compilers with Image File Execution Options and masked off a couple of CPU cores with an affinity mask. Their goal was to keep the shell responsive.
This is one user's anecdote, not a controlled test. It does match Chen's argument that priority alone is a weak lever.
Chen has made this argument before
This isn't a new position. Windows Latest notes that Chen made a similar point in 2010, when someone suggested running Task Manager at real-time priority. Chen's 2010 post argued that real-time priority is dangerous, and that not even input runs at real-time priority, so you can't stop a runaway program interactively if the input thread can't run.
What this means in practice
- A slow Task Manager isn't proof of a scheduling mistake. It can simply be waiting on a shared resource or on the UI system.
- Don't count on a priority tweak as a cure-all. Chen offers none, and the anecdote above suggests the gains are small.
- Focus matters. Chen's advice is that getting focus onto Task Manager or the taskbar is about as good as it gets.
- Developers should take the design lesson. Raising one component's priority doesn't remove its dependencies. Shared locks, slow I/O and UI calls can still stall it. This is my read, and Visual Studio Magazine makes a similar point.
Some readers will find this unsatisfying, because it describes a limit and doesn't offer a workaround. That is accurate to the source. Chen's post is an explanation of a design tradeoff, not a recovery procedure.
In short, giving Task Manager the front of the line doesn't help if the person at the register is still paying for a giant cart of groceries.
References
- Microsoft reveals why the most obvious Windows performance trick doesn't work - Neowin Neowin · 2026-11-16T22:00:01+00:00
- Why don't we run the taskbar and Task Manager at high priority, so they can remain responsive? - The Old New Thing devblogs.microsoft.com
- When you set a 100% CPU program to real-time priority, you get what you asked for - The Old New Thing devblogs.microsoft.com