I agree $35 isn't a ripoff price, but not that 'everybody needs to be paid'.
For instance, Red Gate recently added an F# mode, which essentially doesn't work. If Reflector had been open source and free, I'd have been willing to get this feature running. The benefit I'd derive from a working F# mode would likely make the effort worthwhile.
I could be interested in building a Reflector clone if somebody would write an F# plug-in to the project. I would probably start with looking at the CCI-Metadata project on Codeplex, and the Microsoft Phoenix compiler framework. Mono has similar libraries, but I only know a very small subset of Mono.Cecil API.
I did look into cloning Reflector when Red Gate originally acquired it. The work involved looks similar to that needed to build a .NET compiler, but in reverse. That is, you take a low-level machine representation; then build up a control flow graph; then map that control flow graph onto common high-level constructs such as if statements and for loops.
I'm very familiar with engineering compilers and de-compilers. Have you seen Microsoft Phoenix? It lets you sink/hoist intermediate representations; the key for this would be having an API that allows ordering graph rewriting rules so that you can take the MSIL and convert it the right higher-level form (e.g., should this use LINQ? Well, what does the app settings say the user wants to see the code as?). Anyway, since Reflector is written in C#, it would be able to develop a clone much faster using code generation, since I bet Reflector uses a ton of Visitor patterns and needless C# boilerplate.
I won't promise performance comparable to Redgate's Reflector, but my rule of thumb is Correct. Clear. Fast. In that order.
For instance, Red Gate recently added an F# mode, which essentially doesn't work. If Reflector had been open source and free, I'd have been willing to get this feature running. The benefit I'd derive from a working F# mode would likely make the effort worthwhile.