Grid returned by https://api.weather.gov/points/{point} off for points VERY close to polygon border

[dangelsaurus]

Found an odd scenario where the grid returned by https://api.weather.gov/points/43.1969,-90.4399 actually has the point 1.1 meters outside of the polygon. (calculated using ST_Distance) There are others as well.

Any insight into how the points endpoint calculates the appropriate grid? I know it's probably a bit pedantic, but I'm using ST_Covers to try to find existing cached polygons to cut down on my API calls and these are throwing me curve balls :-)

Thought if I had some insight to how things work behind the scenes, I might be able to program to compensate for these situations better.

[jah1007]

Muscoda, WI sits on the edge of the La Crosse, WI forecast office area of responsibility. The next gridpoint to the east and south are covered by the Milwaukee forecast office. I would expect this behavior as it will use the closest gridpoint for the office to the point.