This fact caused by the qt signal-slot implementation. Even if slots
code is going into parallel they will be called in the main app thread.
And also it is used inside KF5 dataengine implementation, so I'm affraid
that it could not be optimized w\o backend rewriting