IMO, they should. It could be an OS Codeplex project and go from there. Not to replace IE as such, but more like a OS research project. They will need one for Singularity anyway. They probably have one already in research mode. You could make one today using the browser control in .Net. Not sure if the control itself is managed or native.
The web browser control in .NET is just the native code rendering component. If there is a compelling reason to write one in .NET then why not start a project on CodePlex today?
FWIW I think there are advantages to having a webbrowser written entirely in WPF. Not necessecarily because of "security", "plugins" or "speed" but because it would open up the possiblity of applying transforms and other WPF-y effects that you can't do on an interop'd control. That would be a far more compelling argument to me. I'm also not naive enough to believe that such a thing would be trivial (though I'd certainly be interested in helping if there was a project underway!)