It's super buggy! Dupes articles and it's happened twice when long reading sessions were lost without any indication of error - the "read" state of tens, maybe hundreds of articles were not updated. I've seen the code on GitHub in the past - it was a ball of pasta.