So if we're iterating over a list, and outputting a new list that's the same length, that is starting to look a lot like a candidate for map instead of for/fold. Then we don't have to manage an explicit accumulator or worry about manually appending anything.

Consider the following snippet, which is pretty artificial, but incorporates all the same basic moving parts as rdf:li-fixer:

#lang racket

(map (λ (elem)
       (if (> (length elem) 1)
           (apply string-append elem)
           elem))
     (map (λ (num) (list "0" (number->string num)))  (list 1 2 3 4)))

Ah, yes. I like this better.