I've always been an advocate of testing web pages in multiple browsers to make sure everything works.
Now there's new data to support that approach:
As of January, 2009, Safari (regardless of operating system) accounts for 8% of all page views (so one in twelve page requests comes from someone running Safari), and Firefox over 21% (more than one in five).
So if you test only in IE (68%), then it's possible that one in three of your visitors may encounter problems with your site.
Of course, that works the other way as well--don't test only in Firefox or only in Safari. (That tends to be the mistake I make.)
Also, of course, your visitor ratios may vary depending on what your site is about. I imagine that MSDN, for example, has significantly more than 68% IE users.