I agree, but since interface expectations are not written up for Number in general this may not be something that can be fixed quickly.
I kind of expected (one(x) - zero(x)) / oneunit(x) to work for Furlong, but apparently it doesn’t, not sure if that should be reported as an issue.
In general I am not even sure that “removing unit from a number” makes sense for all types which define oneunit but not zero. I am thinking about dates of course, which count from an epoch which should be considered an implementation detail,