My experience has been, though, that many developers have a bit of a laissez-faire attitude towards these kinds of details (time is, of course, another area where precision is incredibly important in corner cases). Rather than expecting all future programmers to be as diligent as they need to be, it's probably easier to simply use tools that make it difficult to make mistakes. Every time I'm working on financial systems I build currency classes (usually I'm doing OOP). Usually people grumble and there will be people who ignore the currency classes. I've never had to wait long before someone trips over an edge cases and everybody understands the benefits.
Ruby and Python have arbitrary precision decimal and rational support in the standard library. While there are arguments tobe had over whether using binary floats by default for simple decimal literals is a good choice, that's a different issue than not having any other kind of numbers available by default.
Ruby, Perl5, Python also fail this. Perl6 uses Rat numbers for things like this and doesn't have this problem.