Beginning with https://codereview.qt-project.org/c/qt/qtbase/+/684038

It makes no sense to me that if we set QGuiApplicationPrivate::lastCursorPosition due to an Enter event (as we clearly should), that it could have an effect on this test. But here we are.

1535: ********* Start testing of tst_QGraphicsProxyWidget *********
FAIL!  : tst_QGraphicsProxyWidget::clickFocus() '!proxy->hasFocus()' returned FALSE. ()
C:\Users\qt\work\qt\qtbase\tests\auto\widgets\graphicsview\qgraphicsproxywidget\tst_qgraphicsproxywidget.cpp(3411) : failure location
PASS   : tst_QGraphicsProxyWidget::windowFrameMargins()
PASS   : tst_QGraphicsProxyWidget::QTBUG_6986_sendMouseEventToAlienWidget()
Totals: 137 passed, 1 failed, 4 skipped, 0 blacklisted, 9881ms

I didn't reproduce the failure on Windows 11, macOS or Linux X11. It only seems to happen on Windows 10 in CI, so far.

https://codereview.qt-project.org/c/qt/qtdeclarative/+/237678 made the assumption that an Enter event does occur. But now we want Qt Quick to rely on QGuiApplicationPrivate::lastCursorPosition (for now, until we move it to QPointingDevice) rather than trying to track mouse position independently.