In Gary Gygax's Gord the Rogue series of novels depict the city of Greyhawk as a great, teeming metropolis, comparable in scale and atmosphere to both Fritz Leiber's Lankhmar and Victorian London as depicted in the novels of Charles Dickens (combined almost certainly with touches of the real-life Depression-era Chicago of Gary's own childhood). In the novel City of Hawks he even included a schematic map of the city showing its various districts and neighborhoods, with a scale that indicated the city covers an area of about 30 square miles - roughly seven miles north to south by 4-5 miles east to west. As I've mentioned previously, Gary describes the city in a way that makes it feel really alive, from the reeking Slums to the glittering High Quarter, with many other areas in between. The city feels almost impossibly, immeasurably large, truly the center of its own self-contained universe. The depiction of the city is by far the best thing about those novels, and it seems almost impossible to me that a D&D player would come away not wanting to use it as a central location in the game (with the fact that it was never detailed in print - and no, the version TSR released post-Gary emphatically does not count! - isn't really an obstacle, precisely because the city is so bi; it can't really be mapped or fully detailed - just use the large-scale district map and at the local level do whatever you want).
And yet, when you look up the entry for Greyhawk City in the World of Greyhawk, you see it listed with a population just over 50,000 (depending on which version of the set you're looking at - it's 53,000 in the original 1980 folio, increased slightly to 58,000 in the 1983 boxed set). That is, to put it mildly, not a teeming metropolis as described in the novels. And that's not just me applying unrealistic modern standards (like it being about half the size of the mid-sized city I grew up in) - in the medieval period there were many cities in Europe and the Near East with populations of 100,000 or more. Even the random city population size table for the World of Greyhawk that Gary included in Dragon #101 generates populations of up to 96,000, which puts Greyhawk's population as just about the middle of the range. In order to plausibly feel like its portrayal in the novels (and fill up those 30-odd square miles of space) Greyhawk City needs a population of at least several hundred thousand people. That the "official" population figure doesn't actually reflect the entire population - leaving out garrisons, criminals, foreign enclaves, and other such marginal groups - isn't enough to make up the difference.
But, having decided to increase the population of Greyhawk City, that creates another issue - if Greyhawk City has a population of around half a million, then the population figures for almost everyplace else in the World of Greyhawk become too low - Greyhawk City shouldn't have a higher population than the entire Kingdom of Furyondy, or the County and Duchy of Urnst combined. Thinking about this and a couple of Google searches showed me that the anomalously low population values of the World of Greyhawk have been a point of contentious discussion in the fan community for a very long time, and in particular a lot of virtual ink was spilled on this topic a few years ago in the D&D blogosphere. So it's not just my imagination - compared to, for instance, Europe in the middle ages, the published population figures are around 5-10x lower than what would be expected, and when we consider that the Flanaess is actually quite a bit larger than continental Europe, the population density is even lower than that, with even "central" civilized regions having about 5 people per square mile (compared to anywhere from 50-100 in medieval Europe).
Those old blog-conversations (and the message-board thread where I brought this up a few weeks ago) offered some justification for those low values - from the "not everybody is counted in those figures" argument again, to various notions of the need for low populations to allow for "adventure-able" frontiers and manageable/wargame-able state-level conflicts, and the notion that in a world that includes both real magic and real monsters the population might stabilize at a low level, further afield to claims of the World of Greyhawk being a de-facto "post-apocalyptic" setting where hard-pressed pockets of civilization are under constant threat of imminent collapse and even the ostensibly-civilized areas are really little more than howling wilderness. While those arguments are all reasonable enough to allow somebody who wants to stick with the published population values to do so, they don't really do it for me. For one thing, because the World of Greyhawk isn't really depicted as a wasteland on the knife's edge of total collapse, for another because most of those arguments and justifications would still apply with a population 3-5x larger (which would still be very low compared to medieval Europe), and - selfishly - because sticking with the "by-the-book" population values doesn't address my original issue: that I want Greyhawk City to have a much larger population, but also don't want to completely throw off the implied balance of the setting.
So, after having given all of this way too much thought (and justified - just like my change to the Oerth calendar - by noting that TSR already increased the population of many states between the 1980 folio and 1983 boxed set, thus undercutting any notion that these values should be treated as "sacred text" or that there was some secret justification behind them that we dare not second-guess), I created a big spreadsheet of the population values for the World of Greyhawk and just semi-arbitrarily increased them across the board. Most countries got their population increased by a factor of 5 over the folio value, some by a factor of 3, a few by less. For towns (population under 10,000) I generally increased their population by 20-40% (to better reflect the range given by Gary in Dragon #101), cities (population over 10,000) were mostly doubled (for the same reason), and the four free cities (Dyvers, Irongate, Rel Astra, and of course Greyhawk) were treated like countries (i.e. population increased 3-5x). Demi-human populations (and human woodland populations) were mostly doubled. The end result of all of this was an increase in total population from 12M to 40M, with Greyhawk City having a population of 265,000. That's still small compared to Victorian London (1M+) but about the same as Paris in the 1400s (i.e. as depicted in Victor Hugo's Notre-Dame de Paris), which I can live with.
I uploaded the spreadsheet, in case anyone wants to see it: https://drive.google.com/file/d/18ARDvoapHXuQkwjhil_BP8928g0j404t/view?usp=sharing
Since these numbers were all derived semi-arbitrarily based on gut feeling (which I'm fairly certain is the exact same way Gary Gygax created the original numbers) I did some further spot-checking, calculating population densities for a handful of areas (states and forests) to see what the numbers looked like and how that would affect the feel of play - whether I'd inadvertently over-populated the place. The results (which I didn't save in a convenient spreadsheet) came out okay - the ostensibly-densely-populated central states having populations of 10-20,000 per 30-mile hex (i.e. about 26 per square mile), borderland-ish states (Geoff, the Iron League states, etc.) having populations around 5,000 per hex (6/sq. mi.), and woodlands around 1,000 (1.3/sq. mi.). Those latter two figures are totally workable, and even the former one is low compared to historical values for Europe (and is comparable to the present-day population density for some mostly-rural counties in northern California that feel anything but crowded). It does mean that the central states need more cities and towns than are depicted on the map alongside the already-assumed hamlets and villages, and that even in the borderland areas every "open" hex will be assumed to have 7 or 8 villages rather than 1 or 2.
That might seem to some people like too much, but I think it's worth keeping in mind that the civilized areas, the open hexes on the map, by-and-large only matter as a backdrop. Characters are from those places, and may be in service to their rulers, and are trying to defend them from the forces of evil, and will pass through them, and may spend their between-adventures time in them, but the actual on-stage adventuring activity almost always takes place elsewhere - in the forest and hill and swamp and mountain and desert and jungle and badlands hexes that surround them - the wilderness areas that remain just as vast and sparsely-populated as they ever were. And even those adventures that do take place in the civilized lands almost by definition aren't going to be exploratory wilderness hex-crawls, but rather will center on investigations and negotiations and other such matters where, except for determining travel time from location A to location B, the map isn't even relevant.
So when traveling through open/civilized areas on the way to and from adventures we can assume that there are almost always roads and inns to spend the night in and no reason to camp outdoors except by choice (if they're outlaws, or trying to keep their presence in the area a secret, or broke), which affects the number of random encounter checks. Having sat through many boring sessions where what seemed like a routine overland trip from point A to point B got bogged down in a seemingly-endless procession of random encounter checks and by the time we got the the ostensible starting-point of the adventure we were already exhausted and ready to call it a day, I see that as a feature, not a bug :)