I’ve seen various answers to how stardates relate to real dates. But is there a general consensus on the correct/most valid answer that holds for all of the various media?
All these quotes are from the Memory Alpha article on Stardates:
The Original Series, the Star Trek Guide:
Pick any combination of four numbers plus a percentage point, use it as your story's stardate. For example, 1313.5 is twelve o'clock noon of one day and 1314.5 would be noon of the next day. Each percentage point is roughly equivalent to one-tenth of one day. The progression of stardates in your script should remain constant but don't worry about whether or not there is a progression from other scripts. Stardates are a mathematical formula which varies depending on location in the galaxy, velocity of travel, and other factors, can vary widely from episode to episode.
Direct from Gene Roddenberry:
When we began making episodes, we would use a star date such as 2317 one week, and then a week later when we made the next episode we would move the star date up to 2942, and so on. Unfortunately, however, the episodes are not aired in the same order in which we filmed them. So we began to get complaints from the viewers, asking, "How come one week the star date is 2891, the next week it's 2337, and then the week after it's 3414?"
In answering these questions, I came up with the statement that "this time system adjusts for shifts in relative time which occur due to the vessel's speed and space warp capability. It has little relationship to Earth's time as we know it. One hour aboard the U.S.S. Enterprise at different times may equal as little as three Earth hours. The star dates specified in the log entry must be computed against the speed of the vessel, the space warp, and its position within our galaxy, in order to give a meaningful reading." Therefore star date would be one thing at one point in the galaxy and something else again at another point in the galaxy.
From the Star Trek: The Next Generation Writer's/Director's Guide:
A stardate is a five-digit number followed by a decimal point and one more digit. Example: "41254.7." The first two digits of the stardate are always "41." The 4 stands for 24th century, the 1 indicates first season. The additional three leading digits will progress unevenly during the course of the season from 000 to 999. The digit following the decimal point is generally regarded as a day counter.
Star Trek 2009 didn't conform to either TOS or TNG-style stardates:
Stardates from the latest film were developed by screenwriters Roberto Orci and Alex Kurtzman. According to Orci, they "used the system where, for example, 2233.45 or whatever means 23rd century, 33rd year of that century, and the .45 indicates the day of the year out of 365 days."  During a Q&A session, Orci restated that a stardate is "the year, as in 2233, with the month and day expressed as a decimal point from .1 to .365 (as in the 365 days of the year)."  A similar reply was posted on his Twitter account: "star date=standard year, with decimal representing day of year from 1-365." 
Orci never said whether leap years end at .366, which would be expected if the digits before the decimal point correspond to Gregorian calendar years, and he didn't explain why stardates 2230.06 and 2233.04 were scripted if .1 is supposed to be the starting decimal. The table below shows stardates from the film.
The page does note that DS9 and VOY used the same system as TNG:
The second digit continued to increase every TV season in other spin-offs as well, even after TNG had ended. Since DS9 premiered during the sixth season of TNG and was set in exactly the same timeframe, stardates on DS9 ranged from 46379.1 to 52861.3. Likewise, the first season of Voyager would've been the eighth season of TNG had it continued, so Voyager stardates ranged from 48315.6 to 54973.4. Star Trek Nemesis, the latest Star Trek story in the 24th century, had a stardate of 56844.9, showing that it took place approximately fifteen years after the first season of TNG.
And finally, throughout the page are comparisons between the Stardate and Gregorian calendars that have popped up in various episodes, notes on inconsistencies, and so on.
First, it has to be acknowledged that the stardates in the Original Series are internally inconsistent. They were simply not managed; nobody thought it would be important. So they don't consistently go up from episode to episode whether you go by production number or airdate. They range from 1312.4 in "Where No Man Has Gone Before" (original pilot but aired third) to 5943.7 in "All Our Yesterdays" (second to last episode). If, for example, one stardate were equal to one day, that would be almost 13 years of in-universe time in our 3 years of TV. And while "Where No Man" was clearly a while before the rest of the first season given the changes between pilot and series, the firmly-in-series episode "Mudd's Women" had the very close stardate of 1329.8.
If we instead assume that the given range is ~ 3 years of in-universe time to match our 3 seasons of TV, that makes a stardate unit equal to about a quarter of a day. Which would put "Mudd's Women" just 4 days after "Where No Man", which seems unlikely given the changes between those episodes.
But those are hardly the worst offenders. Consider the trilogy of "What Are Little Girls Made Of?" (2712.4), Miri (2713.5), and "Dagger of the Mind" (2715.1), which apparently take place over a very short span of time - less than a day, or, at the rate that puts the whole series over 12 years, less than 3 days. That's not even enough time for the action we see in these episodes, much less any travel time between them.
So while the stardates go up too fast over the course of the series, there are multiple places within the series where they don't go up fast enough between episodes.
This is why Michael and Denise Okuda gave up on trying to come up with a consistent formula mapping from TOS stardates to dates in our calendar for the Chronology – it's just not doable without making wild assumptions about multiple unlikely in-universe changes and resets to the system that somehow never came up onscreen.
No reference book can make them make sense; we can quote Roddenberry's line about how stardates are not really meant to form an absolute framework, but everyone in-universe acts as if they do.
In the TNG era, on the other hand, we have much more consistency. Each season's stardates fall within a 1000-unit range, so presumably 1000 stardates is about an Earth year of in-universe time. We also know that the late-season-1 episode "The Neutral Zone", Stardate 41986.0, takes place sometime in the year 2364 on our calendar. (This seems inconsistent with Data's claim to be from the Starfleet Academy class of '78 in "Encounter at Farpoint", but perhaps he was using a different calendar when he said that. All of the explicit onscreen year/century references have someone from the future – e.g Kirk or Picard – addressing someone from our time – e.g. Gillian or the frozen folks in Neutral Zone – and saying some variation of "on your calendar", so maybe an additional calendar, or at least calendar era, is in use by the Federation in the time of TNG; we don't know.)
We also don't know when in 2364 Netural Zone takes place. There were some translations made for on-set decorations and displays, like the Enterprise launch plaque, under the assumption that stardate xx000.0 is January 1st (roughly; I assume stardates are fixed-length and that 1000 of them would be something like a mean Gregorian year of 365.2425 days rather than a variable number of days depending on leap years, but who knows). But if we make that assumption, then "Family" is showing wildly out-of-season weather for French vineyards in January. If you instead assume that each episode takes place in roughly the same time of year that it aired and average out those airdate/stardate mappings, you get stardate xx000.0 falling around July 29th, which seems appropriate as it is Wil Wheaton's birthday. But Discovery S3 broke that alignment; the stardates and years we get in conjunction with each other in the first and second episodes of that season are consistent with the TNG-era mapping, but not with a July 29th start date. They map up if we assume stardate x000.0 is January 1st; we could push it back no further than the latter half of the previous October without changing the year number.
Anyway, from TNG forward, that works. 1000 stardates/year starting with 00000.0 falling somewhere in late 2322 to early 2323.
If you zoom out and let the inconsistent details blur together and just boldly assume the same 1000-stardates/year rate holds for the TOS era, then Kirk's two series - if you toss out one TAS stardate that is way higher than the others and bumps right up against The Motion Picture - covers about 5.5 years, which is pretty close to the stated "five-year mission".
But then the movies don't work; there's supposed to be another five year mission between The Motion Picture and The Wrath of Khan, which are only about 700 stardates apart; that would have to be less than a year or more than 10 years under these assumptions. It just doesn't hang together.
Discovery S1E1 told us that stardate 1207.3 falls on May 11, 2256. If we assume 1000 stardates/year like in TNG and that 4-digit stardates wrap around after 10 years, then the surrounding stardate 0000.0's fall on February 26 (in 2255 and 2265). The latter one at least puts Mudd's Women in the year 2266, matching the Chronology. But it pushes "Where No Man" also into 2266 instead of 2265, and later season-one eps which should also be in 2266-2267 wind up in mid-2268.
Still, it at least gives us a plausible justification for picking a mapping to time of year: we can use it to assign stardate xxx000.0 to Feb 26 (we'll assume it was at midnight UTC at stardate 00000.0 in 2323 and then advanced in subsequent years through the additional fraction of a day in a mean Gregorian year). That would mean "Family" happens in early March, which is probably still a little early but much better than January.