However it's established, the point is that is then has to remain established.
Who thinks that's controversial?
it all depends on how good the DM's notes are, whether prepared before the game or written down during it; because seven real-world years later (which might well only be two or three in-game years for the PCs) when the party go back to that area there's no way in hell I'm otherwise going to remember which town had the good swordsmiths or even if there were any here at all.
<snip>
It's the instances that don't fall under the 'most of the time' banner that cause headaches.
Well, if
no one remembers then it's not going to come up, is it? So there won't be any headaches.
But if someone remembers (whether literally, or by reference to notes) then there are no headaches.
So each player can come up with their own imagined timeline for the setting you're using? That sounds like a recipe for madness.
You're missing my point.
For exactly how many days was Watson stationed in Afghanistan? I'm pretty sure A Study in Scarlet doesn't answer that question. So if the reader is curious, she has to just make it up. Suppose that you imagine it to be some days or weeks more or less than I do - why is that going to cause madness?
In the context of my 4e game, was it summer or winter when the gnoll army killed the king of Nerath around 100 years ago? And was that
exactly 100 years ago, or approximately? And for how long did the empire linger on after the king's death? Nothing in my campaign has answered these question, so each player can envisage it as s/he wants to. What does it matter?
Supppose that an answer were established, it might turn out that one or more of the PCs had made a false assumption. How would that be unrealistic!
The difference being, of course, that in the game your PCs can wander over to those rooms and see for themselves what's in there if the Tower of the Elephant is where they happen to be.
And if they do it can be narrated in the appropriate amount of detail. But there'll always be
something that could have been, but wasn't, investigated. And some detail that could have been, but wasn't, narrated. That's my point. What colour is the timber of the table? Do the chairs have carved or flat saddles? Are the door jams timber or arched stone? Etc.
Even something as simple as an overview map of the continent is bound to include all sorts of campaign (by which I assume you mean setting) elements that the players/PCs might never encounter during play, yet the game-world - and, thus, the game itself - is made richer and deeper by the map's existence. A player can pick the map up, look at it, and let her imagination take over.
I thought you were objecting to the use of imagination in respect of unspecified fiction?
But the map I use in my 4e or BW campaign (I don't have a map for my other games) is not an existing fiction to be explored - it's a list of prompts, or placeholders for possibilities. Which the players know.
So why not just have the setting somewhat heavily developed even if it's the first game there?
My answer would be: there's a big difference between a setting that we've created through play and one that's been invented by one person outside the context of play.