NavList:

I wonder if horizon sights might complicate this a little.  If horizon sights were noise-free, so that it actually made sense to worry about this, then when taking a sight you're sampling the DOV all along the line of sight, which extends for some miles.  (Near Hawaii there ought to be an azimuthal dependence on observed celestial height relative to a distant horizon, more complex than just a constant DOV bias.)  With a theodolite, though, you're just sampling the DOV right at the instrument's bubble or compensator.

Historically, why was there so much emphasis on calculating and standardizing on geodetic coordinates?  Is my feeling right that, back in the day, if you wanted to mark a given location, then return to it later as accurately as possible, astronomic is what you wanted?  Maybe the smoothness of geodetic coordinates was just more appealing; and I guess this issue only ever arose within the bounds of a given triangulation net, where geodetic coordinates could be made to exist, and if you wanted position repeatability without geodetic-coordinate-added-noise contaminating your nice astronomic numbers, you could just drive a stake in the ground, problem solved.

We could, if we wanted, use perfectly spherical coordinates for latitudes and longitudes