The reason appears to be in the qrexec.$VM.log:
Incompatible agent protocol version (remote 3, local 2)
Unable to init server: Could not connect: Connection refused
(zenity:9600): Gtk-WARNING **: cannot open display:
failed
After reviewing what each stage does, I’m left with the impression that stage1 updates VM with 4.1 qrexec-agent, which fails to work with 4.0 qrexec-daemon, and that the 4.1 qrexec-daemon only gets installed in stage4.
But OTOH @marmarek 's announcement states:
If any stage fails, it should be safe to simply retry (potentially after
fixing what it was complaining about).
… but it really looks like that does not hold for stage2, with VMs not being able to start any more.
Is there anything I missed, @fepitre ? Is it safe to proceed after stage2, even though re-running stage2 itself just does not work ATM ?