This means that you have to have code review from the Linux kernel team. And you have to divulge any amateur or buggy code embodied in the source. Which may compromise the imaginary advantage your marketdroids think they have on other platforms.
God yes this. 1000 times this.
One particular example I remember well was TV capture cards in the early/mid 2000s.
Basically the chipset was the Brooktree BT878, which was actually pretty good though remarably cheap. I ended up with a few capture cards what people gave to me because "they didn't work".
That meant they didn't work on Windows. Every manufacturer wrote their own buggy, unstable, system crashy drivers and put effort into some god-awful shiny TV program which made heavy use of gradients and nonstandard TV controls.
On Linux, they all. just. worked. There was one BT878 driver that was well written and well debugged and "shitty" capture cards that "didn't work" gave years of stable, flawless performance.
The same thing cycled around with webcams. It was a wild-west of chipsets. They'd all work after a fashion on Windows. On Linux, they either worked perfectly or not at all due to lack of drivers. The ones that did work were invariable more stable and more featureful because the driver would be written to expose the full functionality of the chipset.
These days the situation is better on all platforms since the standards people have realised that having standard driver interface makes for a much better experience. xHCI means that any random USB chipset works. Same for bluetooth now too. UVC means any camer works and so on and so forth. It's like magic. You can buy a cheap-ass piece of crap from any random vendor and it will just work, no drivers, no hassle on Windows, Linux and OSX.
The thing is vendors are almost uniformly bad at writing drivers. On Linux this means they don't bother. On Windows the drivers are a pile of crap. Having centrally maintained drivers is in fact a large improvement on BOTH operating systems.