This is super interesting! For data quality nerdery: "there's no such thing as a string (without an encoding)", "there's no such thing as a timestamp (without a timezone)", and apparently for geo-data:
There's no such thing as a location (without a relative, timestamped, chain of reference points).
..and if you're doing this RTK-stuff, you kindof need to know that "chain of custody": Here => There => GPS@time
With GPS, we've kindof lost a lot of surveying / map-reading / orienteering in the general population, but looking at the guy trying to map out where the lines in the parking lot are to sub-millimeter accuracy really points out that it's inherently a relative (and time-fixed) process where the local _relative_ positions might not change really appreciably at all, but relative to GPS, over decades there'd probably be some skew (which could be "corrected", but only if you kept that original "chain of custody" of your measurements).
RTK is really about correcting for momentary atmospheric conditions that lead to SLIGHTLY different propagation times for radio waves in order to get much more accurate measurements. The idea is that you take nearby fixed objects that are constantly recording their GPS position, and so over time their precision in their GPS fix goes up. When you hop on with your RTK, it will communicate with those nearby GPS stations and apply a correction based on what they measure vs. the position they know they're actually at.
What you're talking about is totally true though. And it's why defining land boundaries by GPS coordinates is not a good idea.
That's when we get into the wonderful world of coordinate systems. These are deeply involved with RTK, and not knowing about them can mean you get incorrect results even with the RTK, most correction services actually use specific localized coordinate systems, and even offer additional correction data when even that is not enough.
There's no such thing as a location (without a relative, timestamped, chain of reference points).
eg: 38°53′52″N 77°02′11″W (the white house), but needs a timestamp (eg: continental drift, san andreas fault: https://geotripper.blogspot.com/2023/10/why-did-road-cross-s...)
..and if you're doing this RTK-stuff, you kindof need to know that "chain of custody": Here => There => GPS@time
With GPS, we've kindof lost a lot of surveying / map-reading / orienteering in the general population, but looking at the guy trying to map out where the lines in the parking lot are to sub-millimeter accuracy really points out that it's inherently a relative (and time-fixed) process where the local _relative_ positions might not change really appreciably at all, but relative to GPS, over decades there'd probably be some skew (which could be "corrected", but only if you kept that original "chain of custody" of your measurements).