To elaborate on Araq's answer, in Nim like in many other programming languages such as C++ or Java there is a fundamental difference between these two cases.
Your first example shows a regularily defined procedure while the latter does two things at once. It defines a variable and assigns to it a reference to an annonymous procedure. Since the variable is only declared after evaluating the expression to be assign to it recursion is not possible. And this is not the only issue or limitations, for example they cannot participate in procedure overloading (since you cannot have two variables with the same name in the same scope) or dot call syntax. It also gives the compiler less optimisation opportunities since you are dealing with function references now.
