The problem is not mutually recursive modules. Plenty of statically typed languages support mutually recursive modules.
The problem is that it's impossible in general to say what the "interface" of a module is by examining the module alone. This is a very unusual property as real-world programming languages go.
You could fix this by, for example, requiring that all symbols exported from a module have an explicit type annotation. Or, if you think that's not lazy enough, it could be an error to use an imported symbol that doesn't have an explicit type annotation. You could even formalise .hi-boot files if you were truly desperate.
I don't really see this as being any kind of real issue at all. Surely all GHC needs to do is to concatenate all the modules together, alpha-reduce the import/export relations and do a compile/type check over the concatenated module.
If the Haskell spec requires that multiple modules be analysed simultaneously (or multiple times to fixpoint), then that's a bug in the spec. Separate compilation is too important.
Why is it a bug in the spec exactly? And why is separate compilation so important? Chris.