My philosophy is: each paper should have one result (or closely connected collections of results) in it. That is, each paper should tell one story. This is based on past experience: when I put two different (although connected) results in one paper, everybody learned about the most important one, and nobody realized that the paper also contained the second.
For a more famous example, Alexander's paper on the Alexander polynomial contained a bunch of different results, including the skein relation (all connected, of course, because they were all about the Alexander polynomial). Everybody somehow forgot about the skein relation result, and did not realize that this result was also in the paper until after Conway had rediscovered it.
Would it have helped if the skein relation had been published in a different paper? I can't tell ... maybe everyone would have just forgotten entirely about the second paper. And I also can't say for sure why the skein relation result was overlooked. But my theory is that everybody just remembers one story associated with each paper, and that you shouldn't bother putting in anything that's not in that story.
