When to split/merge papers? I ran into the following problem. I have some lengthy paper in which I develop some theory to attack some problem. While I was working on this paper, I got some nice result which might interest a bigger audience and can stand by itself (that is, it is interesting out of context of the big work, and has a fairly short proof). 
The problem: should I write one long paper or two papers, one long and the other short? If I write just one paper, the short result will strengthen the paper, and it will be, perhaps, more whole. On the other hand, for someone that is only interesting in the short one, it will be better if it appears in a separated paper. 
 A: It's hard to give general advice on this, but if there is much work yet to be done on the long paper, I would write up the short one and get it out the door ASAP. If the full paper is almost finished, I couldn't say what to do.
A: If it's not done already I'd say split it.  It's hard for someone else to get motivated to read a long paper, including referees for conferences. Also I have a hard time forcing myself to carefully wordsmith anything longer than 15 pages, and I so get pretty inefficient.
A: According to Gian-Carlo Rota, one of the secrets of mathematical success is to publish the same result many times. 
A: I've split a paper once.  I wasn't inclined to do it, but the paper was becoming huge so I was a bit worried -- 60+ pages, few people read papers that long.  The other criterion was there was a natural way to split the paper into two useful papers.  One turned out to be a "largely survey" 40+ page paper. Sometimes it's tricky getting papers published when they have a lot of survey material, which concerned me at the time.  This allowed for the main result to be a short 20+ page paper. 
In general it comes down to a cost-benefit analysis.  Size of the paper is a consideration.  How interesting the results are, that matters.  Some people go overboard on fragmentation of their papers.  For example, if you're publishing more than 4 papers per year all on essentially the same topic, it leads to confusion among readers as to where your results appear, and which papers depend on which.  
A: My view is that you should always try to write short papers only; a long paper should arise only despite your best efforts to avoid it. In particular, each paper should present only one new idea or theorem (the proof of course might require more than one idea). This benefits both you and your audience. Your audience gains, because it will be easier to both find the right paper to read and read the paper itself. You gain, because you have a longer publication list.
And Rota's advice is quite right. Even one idea or theorem often deserves more than one paper, because there are different angles to present.
A: If I could contribute my penny's worth, people tend not to read very long papers and my experience is that it can often by quite difficult to find somewhere which will publish a very long paper, even if the result is significant.
For example, the two proofs of the Riemannian Penrose inequality (an important result in general relativity) were uploaded to Arxiv at least a good two years before being published in an academic journal, although I don't know if the length of time might have been due to the time needed to verify both proofs, which are very long and technical (the second proof is almost 90 pages long).
