Skymap from narrow posterior samples gives `nan` PROBDENSITY
In some cases, multiorder skymaps from narrow-sky samples contain
In the MWE notebook a set of gaussian samples tightly clustered in 0.01 deg² of the sky is used (and attached here).
I suspect there is something wrong in my usage of either
ligo.skymap.kde.Clustered2DSkyKDE or its
.as_healpix() export method.