It seems this is a win-only bug. I'm fairly sure I know how to fix it, but it obviously won't make it into 0.9 I'm afraid. :-( This is a serious "bummer" since it is a data-loss bug that is so obvious even M$ could see it. Actually, part of the problem in this bug _is_ M$. What happens is that we get WM_DESTROY to the Frame without getting WM_CLOSE, in XAP_Win32Frame::_FrameWndProc that is. I'm quite sure we _do_ get the WM_CLOSE, but it gets routed to somewhere else. I can't fix it in -3.2 hours time, why this will have to make it into 0.9.1. This was something we didn't wish for... /Mike
