If you are considering having a metaplot in your game, reconsider.
First, a definition: A metaplot, for the purposes of this rant, is an onging "story" that is played out in a game world over the course of several supplements, which makes serious changes to the game world in question.
Frankly, metaplots appeal more to people who aren't going to play the game, but either through a collector's urge or due to a novel-reader's desire to "see how the story ends" will (in theory) buy the supplements.
I say "in theory" because while this has worked quite well for some people, and worked okay for others, chances are it won't work at all, and you could have spent your time on something better. (More on that in a minute.)
You have to ask yourself: Do I want my game to be played, or collected?
Because, if you're going to have supplements at all, why have metaplot supplements when you could be doing something else?
And that's an important consideration: Do you need supplements at all? Consider that the need for supplements -- particularly metaplot supplements -- is often more driven by the three-tier system and it's constant, hungry demand for "new product" than by any need of the fans of the game. Consider that you can do quite well by having a complete game, in one book, that is sold direct to consumers over the Internet, skipping the three-teir system entirely.
Given, however, that you are going to have supplements at all, why metaplot supplements? There are any number of things you could produce instead:
The above are just examples. But they have one thing in common: They generally only require the main rulebook to understand, and they are more general than a metaplot supplement.
The main advantage of all these sorts of supplements is that anyone can use these supplements, though tastes and needs vary. Metaplot supplements, on the other hand, benefit only those who wish to follow the metaplot. On top of this, when integrating new players into a game, or when starting a new game, the existance of an officially-published metaplot means that extra time is going to have to be spent saying what the GM does and doesn't use. This is true of any supplement, but it's more egregiously true for metaplot supplements -- it's not just a matter of house rules or what happened in the campaign, but now it's a matter of outlining what didn't happen in the campaign, regarding the metaplot. Why create that sort of hassle for your customers?
If, despite all this, you still want to do a metaplot supplement, be sure to do it right.
Before doing anything, consider having a "setting change" rather than a metaplot. That is, instead of dribbling the changes to the setting out over several supplements in typical metaplot style, consider having it all in one supplement. The comments below still apply, but by putting it all in one place you minimize the bad effect for those who aren't interested (only one supplement is involved), and you can let people know well in advance about the setting changes and what they are.
Regardless, this bears repeating: if you must do a metaplot, please do it right.
First, make sure that the PCs can have an effect on the metaplot, and on the game world. If they can't -- if the metaplot rolls over them, no matter what -- this is frustrating and deprotagonizing. Unless utter helplessness is a major theme of your game, don't do it. And even if utter hopelessness is a theme of the game, consider trusting the GMs to be able to provide utter hopelessness in a better way.
Related to this: When doing a metaplot, don't use it to showcase your favorite important NPCs. This is just like when an obnoxious stranger insists on telling you about his cool character from a previous campaign, only ten times worse. Don't do it. By all means, have important NPCs if you need them, but don't make them the utter focus -- the utter focus should be the PCs. Other people have brought this up before.
Second, go the In Nomine route. Create "Canon Areas of Doubt and Uncertainty". That is, create areas that you explicitly tell everyone (in print) that are not going to be affected by supplements -- and the metaplot. This way, GMs that are interested in the metaplot can run a game, mucking around in those areas, and know that what they're doing isn't going to be invalidated by a subsequent supplement. "Though this is a game about angels, we will never have a supplement that explains the true Nature of God in our universe." "There will never be a supplement on the Inconnu." "The metaplot will never affect the Duchy of Fnord, except insomuch as it affects the whole world. But rest assured that the Duke will always stay in power, even in the roughest times."
This also prevents people who want to be "in line" with the metaplot from waiting for a particular supplement before doing a particular thing with their game and prevents an extreme unwillingness to allow the setting to change at all except insomuch as it does in official supplements. The fault for this annoying form of game paralysis is mainly in the fanboy in question, but for crissakes, let's not encourage it, eh?
Third, provide support for those who don't want to follow the metaplot. It takes a little more effort, but it's be greatly appreciated it. If a customer doesn't like the idea of the World of Kool being invaded by hot pink cockroaches from outer space, at least put in a footnote about what direction the world heads if the hot pink cockroaches DON'T show up.
Hell, consider having several alternative metaplots, which can be switched between depending on what the PCs do -- this links into my first point about doing metaplots right.
Related to this: Do not stuff important rules clarifications/expansions or interesting non-plot-related setting exploration only into a metaplot supplement. We don't want to have to buy "Invasion of the Hot Pink Cockroaches from Outer Space" to get those spacecraft rules for the World of Kool. Sure, the rules for the Pink Cockroaches should be there, but the more general stuff should be available elsewhere.
You can sum up my first three points like this: Flexibility is key. People hate railroading in an actual game, and it's ten times worse when the designer tries to do it from his armchair.
Fourth, connected to the second part: If you say you're not going to do something, don't do it. No one likes someone without integrity. If you say you're not going to do a Hot Pink Cockroaches supplement, don't do one. If you're part of a new team working on an old property, strongly consider the damage that could be done by reneging on the promises of previous developers.
And, since this is a rant, let me leave you with a final comment, though it's been said a thousand times before: "If I wanted to read a novel, I'd buy one." The advantage to an RPG is the story is NOT set in stone. Even in the most railroaded game the PCs can have an effect on the flavor of the game, and most people prefer to have an effect on more than that. Leave the epic stories for the fanfic.
Please note, before you flame, that I'm not saying all metaplots are bad. I'm just saying that before doing a metaplot supplement, consider doing something else instead. And if you're going to do a metaplot supplement, take the effort to do it well. No one is going to get rich in the game industry. If what you're doing isn't a labor of love, is it really worth doing?
[You can see commentary on an earlier version of this, if you like. Many thanks to the people who helped me refine my ravings...]