If we're to look for some sort of a "pseudo-physics" explanation for stardates not really proceeding by X digits per hour, we have to look at the small detail. The big numbers (the first three or later four digits) never pose major problems within TOS, or TNG/DS9/VOY, or even most of the movies. But what comes after the (decimal?) point, and sometimes just before it, is "wobbly" and e.g. responsible for all the cases of overlap in TOS.
Perhaps the 1000 sd/yr assumption is perfectly correct, but the assumption that the things after the dot are fractions of a stardate is incorrect? The new movies also have self-evident pre-dot digits, but nonsensical post-dot ones (say, Kirk is uncertain about the second digit after the dot, which means he doesn't know which day of the week it is, or even which week it is!). Dividing the Earth year in a thousand units already gives you perfectly good accuracy for a calendar, so the things after the dot might be devoted to something else altogether.
We then have to explain why on occasion they do look like fractions of a year (say, in ST:TMP, apparently by design) while on other occasions they do not... And there we can easily go for things like "time zones", i.e. the famous Federation Timebase Beacons ("Clues" et al.) giving instantaneous info on universal time at most locations but suffering from technological limitations and resulting delays at others.
Yet interestingly enough, nobody ever
used stardates to establish the time of day or to count down remaining time, giving us yet another excuse to believe that the things after the dot are not
timekeeping measurements as such. Hours, minutes and seconds are still used exclusively for such applications.