I may be wrong, but I have a vague feeling that it is the other way around: the shorthand is not function, but fun. In other words, function is the basic construction: “function of one argument defined by pattern matching” (remember that in Lambda calculus, all functions have exactly one argument), while fun x y z -> e is a shorthand for function x -> function y -> function z -> e.
In Caml Light (the predecessor of OCaml), the fun keyword could take multiple patterns (like OCaml’s function) as well as multiple arguments (like OCaml’s fun), but this introduced an ambiguity that required inserting parentheses around certain patterns. In the case of a single argument Caml Light also had function (like today’s OCaml), which did not suffer from this ambiguity. Since this issue with the extra parentheses was a bit bothersome, perhaps it was decided not to allow multiple patterns in the fun construct, and instead fall back on fun + match in that case.
@xavierleroy may be able to shed more light on this question.
Cheers,
Nicolas