I’m not sure if you were testing with masks involved but I have seen those
*Float values pop up as well, seemingly randomly. Sometimes they even go negative!
I was expecting RGB values in the range [0-255] but would get very small float values instead.
For me, it seemed to happen when reading color samples when the mask channel was selected in the layer list vs when the “RGB” channel was selected.
So, to work around the issue, I make sure the RGB channel is selected (via BatchPlay) before reading the samples.