The docs state:

Conversion from rational to floating-point respects the following identity for any integral values of a and b, with the exception of the two cases b == 0 and a == 0 && b < 0:

julia> a = 1; b = 2;

julia> isequal(float(a//b), a/b)
true

https://docs.julialang.org/en/v1/manual/complex-and-rational-numbers/

I do not understand the first case (b == 0), because:

julia> isequal(float(1//0),1/0)
true
julia> isequal(float(-1//0),-1/0)
true