In Google's Book Ngram Viewer, prepone shows up in a higher percentage of texts in the early 1800s than in text from the past two decades. However, I don't know how accurate the corpus is for those old texts.
If you look through the documents from that period it looks like they are a mix of words split over lines, non-english words, and ORC errors.
G'Ngrams is very unreliable when it comes to the oldest materials. One clue to that is how the occurrence goes from max to min and back again in less than a decade.
