I'm hacking my way through the underbrush leading into the dark interior of the middle book of a trilogy right now, and finding it heavy going. (Hence the reduced frequency of blogging.)
I find it useful to distinguish between loosely coupled trilogies, and tightly coupled ones. Loosely coupled: three books that can be read separately, but which in combination turn out to revolve around a common theme, and add up to something more than the sum of their parts. Tightly coupled: in the extreme case, it's a single gigantic novel in three acts, so that the second and third books make no sense if read as self-contained entities.
What I'm working on now is probably about 70-80% of the way over to the right on this continuum—a 300,000 word narrative in three chunks, each subsequent one building on its predecessors. This isn't quite the first time I've tackled such a work (arguably books 3-6 of the original Merchant Princes series were intended to be one big fat novel, until a messy collision with bookbinding technology sentenced them to be 300 page episodes), but it's the first time I've gone into the job knowing from the start what it was that I was trying to build. And I'm now at the stage of the project, roughly 40% of the way in, where of course it feels as if I'm making no progress at all and it's doomed to utter futility and incoherence.
One of the problems with writing a novel, or a short story for that matter, is maintaining thematic consistency. Most of the trappings of a work of fiction are modifiable—you can kill off or introduce characters, reverse the plot or reveal the earlier goings-on to be flounderings-in-darkness or misunderstood by the narrative viewpoint, and throw in twists and turns. But the one thing it's hard to mess with is the theme of the story—what it's all about, at the highest level.
Examples of thematic content: "Neptune's Brood" was an abstract exploration of just how one might build a financial infrastructure to support interstellar colonization at slower-than-light speeds, and what can go wrong with it; at the same time, "Neptune's Brood" also asked whether we own money, or money owns us. "Rule 34" was about the future of criminology (not crime) in a networked world, and about how socially acceptable behaviour changes over time. The first Merchant Princes series was, at the highest level, about the annoying fact that actions have consequences, and blood is thicker than water.
The thing is, when you set out to write a book, or a series, you don't necessarily know what your theme is—but it's the invisible direction your compass is pointing you in, and if you lose track of it while you're writing you end up travelling in ever-diminishing circles rather than making progress. Even if you aren't aware of your theme when you start, you'll be intimately familiar with it by the time you finish. If you aren't, what you've got isn't a novel, it's a novel-shaped blob of jelly. And it's generally a bad idea to change thematic track halfway through a book. Your readers will notice the switchback, like the point in the middle of From Dusk til Dawn where it changes gear from a Quentin Tarantino two-psychopaths-on-a-road-trip-with-hostages movie into a completely batshit Robert Rodriguez vampire gorefest, complete with exploding eyeballs and pneumatic-drill-powered wooden stake machine: it's kind of hard to ignore that sort of thing, and it doesn't work well at all unless you know exactly what you're doing and have a palmed card to present to the reader to explain why you're doing it.
To add complications: tightly bound trilogies or really long novels have structural issues all of their own. In a plot-driven novel we usually have a simple progression: from the initial status quo to some source of disruption, then growing tension as the protagonists grapple with whatever's going on, to a climax where the tension is resolved, and a finale that hopefully asserts a new status quo and delivers some kind of sense of closure (closure being what grown-ups want instead of "and they all lived happily ever after"). But if you're writing a 1000 page story that's going to be carved up into 333 page episodes, each of the non-final episodes needs some sort of climax, roughly as powerful as the climax of a 333 page normal novel, but building on each other until we reach the titanic resolution at the end of the story. Structurally it's a sawtooth rather than a single peak in the graph of tension plotted over time.
The challenge in writing a big-ass tightly bound trilogy is that you have to ride a ratchet up a sawtooth of increasing tension with periodic lulls, all the while keeping your eye on the thematic compass. Consequently, writing a 1000 page story that neither sags in the middle nor spirals up its own arse turns out to be more than a little bit harder than writing three stand-alone 333 page stories—even though you can use the same characters and setting throughout.
(And now: tea break's over, it's time to pick up that shovel and go back to work in the word mine ...)