While I'm on my "evil of ..." tour, I thought it might be useful to drop a line or two about the Unidata "deleteuser" command. This command has been purported (by some) to be a "safe" way to logoff a user session. Unfortunately, according to the documentation, this command carries a risk of file corruption (Administering Unidata [7.1] on Unix, pdf pages 173-174) and therefore is anything but "safe".
Now I don't know about your particular bent on this issue, but my personal thought is that anything that is KNOWN to have the potential of file corruption should be avoided inasmuch as possible. Maybe I'm missing something, but discontinuing a user session at the risk of the integrity of precious data isn't a fair trade.
If I need to kill a user session, my first choice is tandem, then UniAdmin. If that's not available or I don't have the credentials for that particular machine, then if it's a Unix variant I'll try a kill with no option on the pid. If that doesn't work, a kill -15 usually does. Neither have the drastic potential of the kill -9 that deleteuser uses.
Certainly, there are times when the kill nor the kill -15 works. In this very rare case, I'll try the -15 several times, and give the machine time to gracefully terminate the session. If after several minutes the session is not being freed, then there may be no other options than the kill -9. In that case, it's likely the situation is so bad that I've already resigned myself to knowing there's going to be some damage to be cleaned up.
But to start with "deleteuser" is, in my opinion, tantamount to taking a sledgehammer to drive a tack into a wall. It may do the job, but can you afford the damage?