As for the https://bugzilla.gnome.org/show_bug.cgi?id=85715, I tried
to #if 0 GTK hack in xterm.c and got clean exit with status 70:

Breakpoint 1, __GI_exit (status=70) at exit.c:100
100   __run_exit_handlers (status, &__exit_funcs, true);
(gdb) bt
#0  __GI_exit (status=70) at exit.c:100
#1  0x000000000056c83d in Fkill_emacs (arg=...) at /home/dima/work/stuff/emacs/trunk/src/emacs.c:1993
#2  0x0000000000510930 in x_connection_closed (dpy=0xc8df90, error_message=0x7fffffffbd20 "Connection lost to X server `:0.0'")
     at /home/dima/work/stuff/emacs/trunk/src/xterm.c:7882
#3  0x0000000000510e83 in x_io_error_quitter (display=0xc8df90) at /home/dima/work/stuff/emacs/trunk/src/xterm.c:7972
#4  0x0000003fe54481be in _XIOError () from /usr/lib64/libX11.so.6
#5  0x0000003fe5445a5d in _XEventsQueued () from /usr/lib64/libX11.so.6
#6  0x0000003fe542533a in XFlush () from /usr/lib64/libX11.so.6
#7  0x000000345c0345d0 in gdk_window_process_all_updates () from /usr/lib64/libgdk-3.so.0
<...skip...>

That's for GTK 3.2.4 on Fedora 16, and the same for 2.24.8. But I read
https://bugzilla.gnome.org/show_bug.cgi?id=85715#c56; can someone check
whether this GTK hack is still actual?

Dmitry