My suggestion would be: Clean up KDE's UI, polish the widgets at the Qt level and do some usability and accessibility testing. Then, remove most of the most redundant options from Konqueror, its menu, its preferences, and Kontrol center's modules and move them to a KConfig panel (a-la GConf or a-la Registry) so both worlds are happy (advanced and newbie users). Only keep visible on panels the most important, basic options, the ones that guarantee a polished experience. If the KDE Project realizes that polish and simplicity is more important than all these almost-never-used and hard-to-find-in-a-chaos-of-panels options, adopt a HIG, get some usability engineers aboard, oh joy, we would have a winner. But as it stands today, the battle with Gnome will still stand as Gnome has an edge on usability direction and general UI polish. I would like it if KDE used Gnome's HIG (which is actually one of the good points of Gnome, so re-use it, that would help the overall usability as a bonus for both DEs and freedesktop.org's efforts).
The fact remains though, KDE is good enough for most people. It does the job. The beta still has bugs and applications crash quite a bit (I think some are Fedora-only issues actually), but it does look very promising. KDE has the right technology and tools. But it is so damned easy to develop for this platform that some people seem to... over-develop.