From the Google Print side, worse scan would probably cause more errors in their behind-the-scenes OCR database linked to each page scan--making searches of these pages less accurate. Hopefully for researches, this increase in error rate will be a fraction of a percent, but who knows?
Once upon a time, I had access to JSTOR, and frequently browsed through their scans of old (as in 17th-18th century) Philofophical Tranfactions. From the glimpses (mostly from the text surrounding my search terms) I got of the underlying OCR text, I came to the conclusion that even error-ridden OCR is good enough to return keyword search results of non-embarrassing calibre. And I can well imagine that some sort of fuzzy term matching to compensate for the most common known scanno themes could be employed to make raw OCR very suitable for keyword searches. -- RS