I see inconsistent failures on the aarch64 CI builder upstream too. But I don't have easy access to an ARM and thought maybe it was unique to my CI.
Looking at these logs more closely, I see a lot of warnings about using g_setenv after threads are created. That sounds like the sort of thing that might cause flaky, architecture-dependent failures. So I reorganized the tests a little to avoid it: https://gitlab.gnome.org/World/deja-dup/-/commit/f82d49984533dc73157ff8dc0038084a4174e8c1 I don't know if that fixes the failures, but maybe? I'd like to fix them too, I've just not been able to reproduce it myself.
