Comment 27 for bug 432964

I guess the issue is hard to reproduce, but we've already 3 duplicates of this. The likelyhood that the system crashes at the very moment useradd is modifying /etc/passwd is so tiny that I immediately thought there was some special casing for gdm-guest-session, e.g. blocking /etc/passwd while the guest session is open. I may well be wrong, though. The other explanation would be that when gdm-guest-session makes the system crash, it often happens at a point where useradd is just starting - this could be improved by waiting for it to return. Just guessing.

The reporting problem is well-known, and on my TODO list, hopefully for Lucid. But it won't be very user-friendly anyways, most people don't know what this means: it took me some time to discover what that lock was. So improving error reporting won't be enough.