i'm working on replacing KUriFilter usage in krunner as that was the major source of slowness that remained. jpwhiting committed changes yesterday to return to the event loop between individual runners during matching to improve responsiveness, and Ryan Bitanga popped up on the panel-devel mailing list to let us now that he has successfully multi-threaded runners and is now working out the remaining bugs ... so we may have a rather snappy little krunner for 4.0.
next on my hit list today is making panels paint their backgrounds properly. this will work very much the way the applet background painting works so shouldn't be too tricky.
one more small step for the workspace, one giant leap for.. well, ok, no giant leaps. =)