I haven't seen split errors in stories for quite some time now, mostly stories older than a few years rather than months. In rare cases I find that a paragraph is not marked with a p but with a h3, don't know how that occurs but it seems more like an error in the uploaded story text than something similar to the occasional split error. A split on paragraphs is simple... if there are recognizable paragraphs in the uploaded text. It's in my line of work to split, aggregate, and correct data files so I know that there will always be situations you can not catch with an automated procedure. Sometimes you have to add specific algorithms for a specific client or a specific type of data file. It won't be different for what Lazeez does with his split code.
I agree that it's better to report it to the author and then here. The exception is the same as one I use for reporting typos to authors: I check if the author has been updating stories in the past. If he has not I don't bother because it's not likely that it will do any good. That goes mostly for older stories so in such a case a conversion error can best be reported here directly and bypass reporting to the author.
