On one hand, I want to remove all that I don't need and keep my privacy on my computer; on the other I don't want to have to fiddle and google why the Windows update failed.
Because, I stopped counting how many times one of those registry changes/app removal actually broke Windows update.
In the end, I just run other software like StartAllBack that I can easily uninstall if it causes issues when updating Windows.
If there is an option in Registry, it's clearly an acceptable use-case.
The truth is that Microsoft developers are lazy (like most devs) and just fail to test their updates properly - or even intentionally ship broken software, in order to force this or that option back to its (typically privacy-invasive) default.
"I deleted a whole bunch of stuff in xorg.conf and my graphics started flickering" is not the fault of Xorg needing more tests.
If the problem is that Windows Update cannot work unless all patches apply flawlessly, then that's a failure of design.
If the problem is that the patches themselves assume that things they provide as options might not be set how they wish them to be set, then that's a failure of testing.
Not at all, it's better for an update to just immediately fail due to unsupported changes made to the OS than potentially leave the system in a unstable state.
Windows was going to move to an image-based OS where updates would be atomic, but sadly that never happened (Windows 10X)
¹: Yes, that’s hyperbole.