Multiple NS records, by design, are load balanced. Multiple A records
are load balanced.
Why shouldn't multiple MX records be load balanced? What makes them so
special to be treated differently?
All you get in either case is a very crude form of loadbalancing. It's not
like you'll see: "MX A is quite busy now, try MX B" and so forth all the
way down the chain. All RFC 1123 stipulates is that the sending MTA should
pick an MX at random, if you have multiple MX RRs with the same
preference. So, your MTA might hit busy ones more times than not.
If you want to do that sort of thing, it would probably make more sense to
have e.g. four MX with the same preference, and use Round Robin DNS load
balancing with A records for the MX'es.
That said, you would've thought that a big, wealthy organisation like
Hotmail would be able to use a proper load balancing front-end to their
MXes, like UU.NET does.