Perhaps it is you that are mistaken. In terms of OS research, I'm aware of Singularity and Midori, but these were all flawed projects that were subsequently discontinued.
No, they were projects killed by management, they were only flawed on the eyes of MS haters.
They had a huge impact on .NET AOT compiler for Windows 8, .NET Native for UWP, Linux subsystem, Windows containers and secure kernel. Including asynch/await, .NET TPL, concepts now spread to other languages.
You are missing Drawbdrige, Singularity's sucessor, where pico-processes where researched and are the basis of the Linux subsystem.
Or the P language, a type dependent language, used to write the new USB stack.
And many other findings, easy to track down at their website.
Well, of course they were killed my management. And they were killed because their performance was so bad that it wasn't worth investing additional time, resources and money into. As for the impact of their research on AOT compilers in their runtimes, what exactly did they revolutionize in AOT compiler design or technology that wasn't already known?
I think you have mistaken Google for Microsoft Research, in what concerns pushing OS research forward.