About 25% of ZIP codes don't have a corresponding Census Bureau ZCTA, for example 10118. Do you end up needing special handling for those cases? Or has it not yet come up in practice?
But I'm not a fan of these particular maps because the use of 3d makes them harder to read. The isometric view and rotation away from north at the top break conventions that people use to orient themselves in the map and connect it to their lived experiences on the ground. I'm reasonably familiar with NYC geography, and I could not immediately recognize the landscape I was looking at in these maps. Ironically, it was only because I already knew the answer to the question that I could do so: "oh that huge green spike must be Manhattan".
I think a 2d choropleth map with a diverging color scale centered on the mean value would work better.
At 1km resolution, this is largely just a map of US urbanized areas. The spatial variation isn't that interesting. The temporal variation and the absolute values per spatial unit are the greater value of this project.
I think adding something like (or awareness of datum transformations generally) would make the tool harder to use for its intended audience. If you need the level of precision where tectonic plate movement matters, then you need professional level tools.
My experience has been that data generalists should stay away from geospatial analysis precisely because they lack a full appreciation of the importance of spatial references. I've seen people fail at this task in so many ways. From "I don't need a library to reproject, I'll just use a haversine function" to "I'll just do a spatial join of these address points in WGS84 to these parcels in NAD27" to "these North Korean missiles aren't a threat because according to this map using a Mercator projection, we are out of range."
DuckDB is great, but the fact that it makes it easier for data generalists to make mistakes with geospatial data is mark against it, not in its favor.
https://docs.moodle.org/502/en/Using_certainty-based_marking