When Microsoft, along with HP and Intel, offered the C# Standard to ECMA in 2000, it was a major break in tradition. Up until that point, all programming languages developed by Microsoft were controlled solely in-house. While documentation was certainly provided, the general populace never got the in-depth to really explore the nooks and crannies of the language, let alone write their own compiler.
An unfortunate side effect of being a standard is that it is written like one. Often the standard will say a compiler may do something but is not required to. While this is fine at a conceptual level, it makes it harder than necessary to what the Microsoft C# compiler actually does.
This where the book by Jagger, Perry, and Sestoft comes into play. They take the C# 2.0 standard and fill in the blanks. This book is full of detailed information on how the Microsoft choose to implement its C# compiler. And though possibly dated, it also includes the same information on Mono's C# compiler.
The real value of this book comes from its discussion of edge cases. For example, the standard includes a section on the FGAB problem. Specifically if F(G<A, B>(7)) should be interpreted as F having the arguments G<A, B> and 7 or the arguments (G<A) and (B>7). While not much of an issue in C# 2, this far more important now that passing functions to other functions is becoming common. The authors build on the standard both by clarifying the rather esoteric text and by describing how the compilers actually handle it. (Both Microsoft and Mono simply fail to parse it in this case.)
If any complaints are to be made about this book, it is that it tends to be very dry and technical. While they do include a lot of background information about how the language came to be, it never quite gets to the conversational level we find in Framework Design Guidelines.
Annotated C# Standard by Jon Jagger, Nigel Perry, and Peter Sestoft is currently in print and also available via Google Book Search.