New arXiv procedures? Recently I encountered a new phenomenon when I tried to submit a paper to arXiv. The paper was an erratum to another, already published, paper and will be published separately. I got a message from arXiv saying that I need to join the erratum with the original file. I was a little surprised receiving a reply from, obviously, a human being. Although I thought the request was a bit silly, I did what was requested, submitted the joint paper (the original union the errata), and forgot about it. But today I got a call from another mathematician. She tried to submit a paper with a title  "... II". The paper "... I" was already in the arXiv and submitted to a (very good) journal. Both papers solve similar but different problems. One of these problems is at least 40 years old. Her submission was denied:  she got a request from the arXiv to submit a union of that new paper and the old paper instead. This is quite silly. Is there now a special person in the arXiv who is making these decisions? It looks like there has been a change in how arXiv is managed. I understand that this is not a research question, and I make it a community Wiki. I post it here because several frequent MO users are affiliated with arXiv. 
 A: I'm still the chair of the math arXiv advisory committee, which admittedly hasn't done a whole lot lately, and one of the global math moderators.  No, there has not been any dramatic change in the management of the arXiv at Cornell.  If anything, I wish that by now more might have changed.  The arXiv has always had the bare minimum funding, sometimes less than the bare minimum.  They have never had polished public relations to properly explain small changes in policy.  (Actually even wealthy Internet companies sometimes stir up confusion when they make changes.)
At some informal level, they/we have always worried about duplicate submissions, and near duplicates, and errata posted as new papers.  And yes there is a new text overlap tool to detect both plagiarism and self-plagiarism.  There is no good, rigorous way to draw the line for any of these issues.  (Just as there isn't at MathOverflow --- what exactly is an "exact duplicate" of a previous question?)  Regardless, if your submission is rejected, you do have the right to "file" an appeal with the Cornell staff.  If it is a plausibly sane appeal, then they should show it to the math moderators and/or the math advisory committee, more likely the former these days.
One perfectly valid consideration is to have the arXiv correspond to what is published in journals.  Although there are cases where strict adherence to that rule is untenable.  For instance, my mother and I have a joint paper in the Annals of Mathematics that appeared twice just because the first time, the paper had TeX symbol encoding errors.
Also, I personally think that this posting is reasonable for MathOverflow.  However, it would have been better with a less suspecting tone.  The arXiv doesn't always make the best impression, but long-time users know that actually it has gotten better over the years.  For a long time it had a reputation as a "user belligerent" web site.  Even then, it was still a force for good, obviously.
A: I talked to the arXiv staff about Olga Kharlampovich's submissions and I now have some answers.  The letter that Olga posted here is a form letter that doesn't fit the facts.  The text overlap tool reported that the new submission substantially overlapped with the old submission.  After that, as far as I know, no moderator and no advisory committee was ever contacted.  Instead, an arXiv employee sent this stock response just to keep things moving.  After that, I was told, her case was added to the to-do list.  I was assured that as of last week, before this question was posted to MathOverflow, her submission was already slated to be reverted in her favor on Monday.
Obviously this is not satisfactory.  I am one of the moderators (and not the only one) who should have seen the appeal.   The e-mail said that someone like me had seen it and rejected her appeal, but apparently no such thing happened.  It seems that the submitted version (which I think is now version 3) had something like 75% text overlap with the previous version (version 2) of arXiv:1111.0577.  It's not so unreasonable to flag such a submission.  After that it wasn't handled properly.  I do not want to name names and lead people to pour opprobrium on the overworked arXiv staff.  (There are only two of them who handle daily submissions.)  But I want to make this story sound accountable, so I can say that some of my information came directly from Paul Ginsparg.
To go back to the title question, no there has not been any great change in arXiv management.  You could certainly argue that there is insufficient management, but that's not the same thing.
People are also asking about the policy by which papers are labelled as having text overlap with other papers.  A clearer statement of that policy would be useful, but that is a separate question from Olga's case.

According to e-mail that I just saw, this morning Olga was given the option of reverting the previous arXiv paper to Part I and submitting Part II separately.  Her answer, according to what I saw, was that she elected to keep it as a replacement after all.  I am mentioning this so that readers who see arXiv postings this week won't think that injustice continues.
I stand by my explanation that the stock e-mail that she was sent didn't fit the facts, and that her appeal should not have been stonewalled.  (In fact her appeal was soon seriously considered internally, but that was not explained.)  However, in the original posting, Olga's name was withheld supposedly to protect her interests.  Although I understand that anonymity is sometimes vital even in a public accusation, in this case I don't see how it helped matters.
A: Hello,
I have paper 1 in the arxiv (that is submitted to the journal) and submitted paper 2 with completely new results (with similar formulations and refereeng to paper 1. I didn't want to change paper 1 because it is submitted, people refer to it, and it makes bad impression when new and new revisions are made, also the submission date is changed), the second paper was returned by the arxiv, I appealed, and this is their response:
Dear Olga Kharlampovich,
Our moderators have considered your appeal and maintain that your article is not appropriate as a new submission to arXiv. The new ideas should be incorporated into a replacement of your existing article.
In general the maintainers of arXiv choose to exercise very limited control over submissions; however, we do want arXiv to be as useful as possible for all of the various communities publishing here.
A moderator noticed that you have submitted several articles in a short period with similar ideas and content.  After a discussion of your submissions among the other moderators and members of the advisory committee, we have decided to ask you to consolidate articles with similar content, or which are variations on the same theme into single articles.
This will be more efficient for the whole arXiv community, and may be beneficial to you as well. In consolidating your work you may find that you can more clearly elucidate the connections and expose the underlying principles so that your ideas will be more useful to others.
--
arXiv moderation

Let me add, that "several articles in a short period " were these Article 1 and 2". The first one was submitted in the Fall, and the second in May. I incorporated them into the same article now, but I think this is silly. What is going to happen if we get new results on a similar topic?
A: This is likely unrelated to the changes that Mark and Will noticed, but the other day a novel
(to me) 
arXiv admin note (under Comments) caught my eye:


A: Let me add, that "several articles in a short period " were these Article 1 and 2". The first one was submitted in the Fall, and the second in May. I incorporated them into the same article now, but I think this is silly. What is going to happen if we get new results on a similar topic?
