Sometimes, while I'm reading a book I start wondering what an author was thinking about when he started writing the novel. Since it's pretty rare that an author actually tells us, I'm in the happy position of not easily being shown to be wrong...
A case in point is Mervyn Peake's Titus Groan. The books starts as if it's going to be a sort of other-worldly Dickensian novel. We get a swooping overview of Gormenghast Castle, filled with the sorts of grotesque caricatures that Dickens liked to write, down to the names (Dr. Prunesquallor, the cook Swelter and his rival Flay, Lord Groan, and so on). And so I wonder if Peake had originally intended for the whole book to be a satire, but then realized that satirizing a non-existent place was too limiting for his imagination.
Because, as the novel unfolds, the characters acquire depth and pathos -- there's a sort of turning-point half-way through, when the castle library burns, that brings out the best and worst of these characters. And it seems that Peake's attitude toward his characters changes then as well -- up until that point, the authorial voice is pretty cold, leading us to despise most of the characters, but afterwards, we end up feeling more sorry for them, or impressed by them, as the case may be. And I think it's interesting that the only characters who don't come out as more well-rounded are the twins and Swelter, who are also the characters who aren't at the library during the fire.
Another example would be Reginald Hill's Bones and Silence, which feels like a really great short story grafted on to a mediocre novel. The novel is about Dalziel's attempt to prove that a man killed his wife, where all the evidence points to suicide. By the end of the novel, he's linked the man to four murders, which would be OK, but there's just so much of the hugger-mugger that gives detective novels a bad name -- the obscure clues, the convoluted motivations, and so on. Interspersed through the novel, Dalziel receives letters from a suicidal woman which he refuses to take seriously, and passes on to Pascoe. Pascoe's attempts to unravel the letter-writer's identity make a great little story, but they're too split up by the main novel to be truly effective.
On the other hand, that story only works because Pascoe's got a number of characters from the main story to draw from -- in a short story, it'd be obvious too quickly to us the readers who the letter-writer is. So it seems like Hill had this great story in mind, but couldn't find a way to make it stand effectively on its own. He needed to introduce more characters, and we end up with all the business of Dalziel playing God in the York Mystery Cycle. It's great stuff, but too flimsy to hang a novel on, and I can imagine Hill just deciding he didn't want to throw it all away, and sticking the other story in on top. Unfortunately, in the end the weaker story dominates the novel, which is a pity.