The GeoPointFieldMapperTests has had dozens of improvements over the last couple of years, while PointFieldMapperTests has had only a few. Some of this is likely due to deliberate decisions, like the addition of Synthetic source support to GeoPointFieldMapper, but not to PointFieldMapper, but others due to prioritising geo_point over cartesian point.

When working on the support of STATS centroid=st_centroid(localtion) in #104218, we likewise did not improve the PointFieldMapperTests to cover the fact that this new ES|QL function supports both geo and cartesian points. The ES|QL integration tests thoroughly covered both types, so the behaviour is well tested, but getting the unit tests updated would be a good idea. The reason it was not done as part of the ES|QL work was because the testing in GeoPointFieldMapperTests relied on the framework used for synthetic source support, which is entirely missing in PointFieldMapperTests. Perhaps we need to add synthetic source support to the same level as geo_point to make this easier?