Comment 1 for bug 1938325

Yes I guess this is a race where the symlink is not in place by the time the copy runs. It's a bit hard to be believe but it looks like this case threads its way through the casper code without hitting a udevadm settle, so presumably adding one somewhere would make it work. OTOH it can take a surprisingly long time for devices to show up at all (this is why there is a loop around the calls to find_livefs) so maybe we should have a loop waiting for ${live_dest} to appear too.