Fwiw this will also let the mouhid.sys driver send WM_POINTER* messages to the input desktop. As it's a service it has a different desktop, and will otherwise try to find a target window in its own desktop.
Right now, only rawinput messages are able to cross the desktop boundaries as we enumerate every process for them, but this is not correctly done according to tests and this MR is meant to fix that.
Also, the pointer messages are not rawinput messages -their rawinput HID reports are sent separately-, so we would have to add another process enumeration for them and I think implementing the input desktop properly is better.