I am saying that for a lot of applications Winforms makes alot more sense, and like-wise WPF makes sense for alot of other applications. I have worked on several lsarge scale WPF applications in recent years, and all of them have been hybrid applications where a control worked best in some situations that was made in Winforms.
In agree in your assesment in the slowness of WPF starting up, but the ability for developers to change an application using data templates and the flexible layout system, has resulted in applications that are pretty much impossible to do in Winforms. It is quite surprising what a developer can come up with in an iteration with some SCRUM pressure, where they would have been left scratching thier heads in winforms owner drawing conrols.
I am currently working on a hybrid WPF application, and the WPF part always takes at least half as long to do as the winforms part. It is a high performance scientific instrument with a lot of precison and math and WPF has really shone. My only regret is the lack of mature open or commercial source controls (graphs at the moment that can render 1000's of points without lag), otherwise we would have gone for a complete WPF application.
I also have a few graduates that are finding it far easier to grasp and use Winforms as opposed to WPF as conceptually the hurdles are just too high for them at present. It will take time for people to come round fully to the merits of WPF, but they will, and peoples issues will be dealt with though I find it hard to see a reason to update our current application to Visual Studio version next as we have a pretty complete toolkit. Time will be taken to optimise things, but I have not had the horror stories with speed on the whole because you can always reduce load up speed with some smart coding