In gtk/gail, the widget that last had focus before its parent top-level window was deactivated always fire focus events when the top-level window is reactivated. Firefox currently only fires focus events on window:activate if the focus changed programmatically while the window was inactive. In sticking with the norm on GNOME, Firefox should probably fire focus events for controls after all window:activate events. If not, an AT will have to predict whether a window:activate will be followed by a focus: or not, and decide whether to use its own cached focus point of regard or wait for the focus event to occur.

Is this a dupe of bug 368322 ?

Marking as dupe of newer bug which has more info.