Problem with FSS verification

I am using verify_fuzzy() to calculate FSS metrics for different case studies.

I works perfectly but I've just noticed that when I run different times, I do not get the same scores always. The FSS scores change every time I run the script. I have verified that the model and the observation grids are always the same, but the FSS output does not.

Is anybody experiencing this?