A fascinating NY Times piece on comparing versions of Beethoven's works:
The problem with establishing a Beethoven urtext is that the original sources are so many, varied and conflicting. Beethoven would usually begin by sketching out a sonata, either in a book or on loose pages that tended to separate, and write so close to illegibly that only a practiced eye could make sense of it. “I served a long apprenticeship deciphering Beethoven sketches,” Professor Cooper said.
An autograph manuscript in Beethoven’s hand would be followed by a fair copy made by an assistant to send to the publisher. The publisher would produce a first impression (a trial run of, say, 100 copies), then a second impression of many more. After that — the world of 18th- and early-19th-century publishing being cowboy country — it was not uncommon for rival publishers to issue their own editions, sometimes at Beethoven’s request. In other words, he double-sold, even triple-sold, his work.
At every stage in this chain came opportunities to change the text, either by accident or by design. But the difference between an error and a correction or improvement is not always clear, so you can end up with five or more variants of the same text with no conclusive proof of which one represents finality.