The advantages would be twofold. First off, people who need to use a 10 year old version of Internet Explorer because of crappy internal apps could continue to do so, and they could install and use a new version of Internet Explorer for using the real world, modern Internet. Secondly, we'd avoid having to hear about the (legitimate) woes of developers who need to run multiple copies of Windows virtually to do testing, or use third-party hacks to get multiple versions of IE to run on a single copy of Windows.
And I'm suggesting that the above be done for new versions of IE. But it's just as possible for Microsoft to release an application version of IE6 that users more recent versions of Windows could install in order to run those ancient internal web applications. There's benefit here too in terms of showing people that they need to use an old-and-busted version of a program to access their internal stuff vs the new-hotness they'd have to use to access the public Internet. Even rebranding Internet Explorer with a different name would help toward that somewhat (if you could install it along side the ancient version of IE6).
Similarly, I hate that Chrome only allows one installation, with the exception of Canary (windows only). What's the point? Give them a unique ID, and namespace everything! Get clever with hardlinking binaries if nothing's different, but sheesh.
</rant>
That's likely the best solution, yes. I guess we can hope it goes that way eventually, right?
Enforcing updates would also likely improve IE's market share, as all those not-recent versions are part of the cause of developers migrating towards Firefox / WebKit browsers. How much has this hurt Microsoft? Enough to make it worth losing the stragglers who aren't paying to update anyway?
edit: of course, there's a solution to all this: the ability to specify the particular version of a browser you're targeting, if it matters. Then at least a warning can be thrown, and if multiple renderers are included, they can be used selectively.