In "Writing with Two Heads," the chapter on collaboration, he says:
Here's [Donald E.] Westlake's description of the process: "First we sat down and discussed the whole thing at length. Then I wrote a fifteen-page outline of what we had discussed. I gave this to Brian [Garfield], and he expanded it to forty pages, putting in all the historical context and everything. Then he gave it back to me and I cut it back down to twenty-five pages. At this point we were thinking screenplay, and this version was shown around as a treatment. When it didn't fly, we decided to do it as a novel first.
"I wrote the first draft, limiting myself to action and dialogue -- not where they were or what they were wearing, just what they did and said. My draft ran about thirty thousand words. I gave it to Brian and he doubled it, turning each of my pages into two pages, putting in all the background and such. Then he gave me his sixty-thousand-word version and I edited it, and I gave it back to him and he edited it, and then we gave the whole mess to an editor."
"It sounds," I ventured, "like five times as much work as sitting down and writing a book."
"Yes," he agreed, "and about a quarter as much fun, and for half the money."
I've never collaborated with anyone on anything I/we intended to publish. When you're focused on the product, and thinking about how readers will like it and react to it, it does seem like the process could become rather fraught, and tempers might flare. Any writer I'd care to collaborate with is someone I'd like to remain friends with, you know?
I've done collaborations for fun, where about ninety percent of the point was the process rather than the product. Another word for this is "playing," and you can do it with two or more people in a chat room, or e-mailing each other. Everyone is playing a character, and you go back and forth, typing what your character is saying/doing. (Yes, it's pretty much exactly like kids playing Batman or cowboys or cops-n-robbers, although hopefully the folks in the chat room write better than eight-year-olds.) With the right people, and the right characters and set-up, this can be a blast and a half. I've participated in this kind of collaborative writing in stories that went on for years. We posted them online as we went and hardly anyone read them, but that wasn't the point. Hanging with friends and having fun developing characters and creating story was the point, and the fact that hardly anything produced this way is worth publishing commercially is completely irrelevant.
Some writers collaborate a lot, and they've clearly figured out a way to make the product worth whatever aggravation the process causes. Or maybe they've worked out a process is smooth and efficient, in which case I'd love to read about it.