appears to be in the updateSource() method of the time engine. It uses a different logic to sourceRequested(). Attached is a patch to do the same as in sourceRequested(). I still can't duplicate the problem, even if I place a second clock on the desktop and wait for the time to change. Is there any chance you could compile kdesupport, kdelibs, kdepimlibs and kdebase with this patch and test it again?