Among old school fans, it's a popular sport to make fun of the Forgotten Realms, something I've never entirely understood. While it's certainly true that many of the products and novels published by TSR and WotC made the Realms look absurd, for me, it was Ed Greenwood's articles in Dragon that most fully depict the setting. I remember reading many of those articles with great interest back in the day and thinking how wonderfully evocative they were.
When TSR published the Forgotten Realms boxed campaign setting, I was quick to purchase a copy. What I remember most clearly about reading it for the first time was how much more detailed were the Dalelands and Waterdeep and surrounds compared to the rest of the setting. Indeed, the Realms struck me as strangely undetailed outside of a few areas and I wondered why at the time. In retrospect, the answer is obvious: Greenwood only detailed those parts of the Realms that he had to as his campaign progressed. The rest was left vague or unexplained, to be picked up later when and if it ever became necessary. That's more or less the exact approach I've adopted in my Dwimmermount campaign, where anything "off the map" remains, for all intents and purposes, unknown, even to me.
What's interesting is that, once upon a time, this was the norm. Dave Arneson's Blackmoor setting, despite being its age, is remarkably limited in its scope, centered as it is on the kingdom of Blackmoor and the territory immediately surrounding it. Information about anything beyond that comparatively small area is sparse, undoubtedly because such information was unneeded in play. The same seems to have been true of Hargrave's Arduin and (initially anyway) Stafford's Glorantha. Conversely, Barker's Tékumel seems to have had a lot of thought put into it outside of the context of actual play in the setting, with Professor Barker going so far as to create extensive socio-cultural (and linguistic) details for places that had little or no impact on the campaigns he refereed.
I'm not quite sure where to place Gygax's Greyhawk setting. Its published form has some relationship to actual play in Lake Geneva, but the relationship is not one-for-one. The World of Greyhawk owes its existence at least in part for a need on the part of TSR to present a sample campaign setting for D&D players to purchase. Most of the later development of the setting has absolutely nothing to do with actual play in Lake Geneva, being created solely to sell products. Dragonlance's Krynn is even more a creature of marketing, being a setting designed by a committee to launch a new brand. Unless I am mistaken, the only parts of it that derive from actual play by anyone involved are its deities, which were inspired by those in Jeff Grubb's OD&D campaign, as outlined in his The Matter of Theology. TSR's 2e era settings have even less grounding in actual play.
I sometimes come across as being opposed to the very idea of published campaign settings and that's not true at all. I have great fondness for many of them and have often used many of them over the years. What I actually object to are settings that don't reflect anyone's actually having used them in campaign play. I dislike settings that exist solely as commercial products without a foundation in play. The early to mid-1990s were therefore to my mind an awful time in the history of Dungeons & Dragons, dominated by the publication of setting after setting conceived from start to finish purely as commercial products and nothing more. So, it's not that I object to publishing campaign settings as such, but it'd be nice if those settings had a history deeper than the project pitch made to fill a hole in a release schedule.