In standard FORTRAN or Fortran, the operands of a boolean expression can be evaluated in any order. Incomplete evaluation is permitted, but implementation defined.
This allows optimisation of boolean expressions that would not be permitted if strict Left-To-Right ordering was enforced. Expressions which require strict ordering must be decomposed into seperate conditionals, or implementation-dependent assumptions can be made.
Since decomposition is used to enforce ordering, it follows that seperate IF statements can not always be optimised into a single expression. However, short-circuit evaluation is explicit with decomposition, and this is never worse than languages which enforce strict left-to-right ordering to allow lazy evaluation.
Languages wich are derived from FORTRAN (Fortran, BASIC, VBn), and languages which were designed to achieve FORTRAN-like efficiency (Pascal, Ada) initially followed the FORTRAN example of allowing out-of-order evaluation.