Losing tiny slices of floating point somewhere

[zenna]

I think I am losing floating points in the refinement or somewhere.

e.g.

prob(flipsmt(0.4)) => [0.3999999999999986 0.3999999999999986]

One way to see what is going on would be to hijack call(X::RandVar{Bool},...) so that it always returns TF. Then see what probability bounds I get. I suspect the answer will be [0,0.9999] something, which is wrong.