At least xkey.state is ignored and a real (hardware) pressed (mainly "control") modifier skips several keys completely.

Also see
https://www.mail-archive.com/interest@qt-project.org/msg12085.html
https://bugs.kde.org/show_bug.cgi?id=353781

To a certain degree, this can probably be circumvented by using the xtest extension, but nevertheless, this seems like a bug in Qt which may hit if the logical state of an event differs from wherever Qt takes the modifier states (libinput?)

If synthetic events are not meant to be supported, they should at least be dropped altogether and not be interpreted differently