The fact you merely once did tech support and then moved to PHP development explains why you hold ridiculous opinions on various things.
I'm sure Winsock reset _did_ solve a lot of problems, for you. But at the same time you probably just made other problems for the poor peoples affected by your tech support. Whilst Winsock reset may have removed the particular third-party component that was causing the issue (often shitty AV software), it will have removed any others that had valid reason to be there, such as VPN software. Do some reading about LSPs (layered service providers) in the context of Winsock and educate yourself.
"Crappy Win32 APIs"
Citation needed. (Figure of speech on HN, don't worry I'm not expecting you to actually try to find one)
"Somehow in your decades of using Windows you have never run into buggy, crash prone applications."
Don't put words in my mouth. Of course I've experienced buggy crash prone applications. Just as I've experienced such things on OSX and Linux. As long as such shitty apps don't bring down the OS or your Shell, all's fine really. Don't blame the OS for shitty third party crap.
"And it happens often enough that if you're in any sort of support role,"
Starting to see a pattern here. You misdiagnosed a lot of problems when you used to work in tech support, huh? Don't blame you, gets them off the phone doesn't it?
Nobody denies the Registry is stupid by design. But that doesn't make it buggy and broken like you claim. Microsoft acknowledges the Registry should never have happened and they do their best now to try to encourage better practices.
"I can't find any better sources for why I thought that the driver model itself was wrong,"
Because you won't find any but don't let me stop your desperate web searching to back up bullshit claims.
"but I do remember that a lot of Windows crashes happen because of crappy drivers"
Well thanks for that Mr Captain Obvious. Just like shitty drivers on ANY of the big 3 OSes can and do cause crashes? With the other big percentage of Windows crashes being faulty hardware. Fortunately Windows has advanced itself to such a point now where the vast majority of drivers, especially those most likely to be badly written (i.e. all things USB pretty much!) are held as user-mode processes that are unable to crash the system.
"I remember there was a specific reason for it, and it wasn't fixed in as late as Windows 7, and it's even considered a feature."
Nothing was "fixed" in Windows 7 concerning the fundamental stability of the driver model in Windows. They make subtle and often completely unnoticed resiliency improvements all the time. NT 6.0 (Vista to laymen) added the capability to automatically restart a kernel-mode graphics driver. Which happened to be quite a valuable feature at the time because Nvidia especially was really struggling to write stable LDDM (aka WDDM now, the L stood for Longhorn) graphics drivers at the time. It took them a good 6 months to sort those issues, during which time they were leaning on the inherent resiliency of the NT 6.0 kernel to keep customers sane.