1) In testing our new cluster with OSX and AFP volumes, we see that an
abnormal disconnect from the server (like a dropped VPN connection) will
leave the user connected to the server, but when they try to reconnect,
it refuses them (because in our case we restrict users to 1 login).
This used to happen a lot with IPX (less with IP), especially with an
abnormal termination (read Windows crashed).

Is there a timeout that will eventually drop this connection, or will we
have to go in an manually clear the connection? I'm really close to
giving everyone two connections just to avoid this, but we've
historically not done this because of billing and licensing issues.

2) We also saw a strange problem with launching a vpn client when the
user was already connected to the server. Obviously, this is a no-no,
but it can happen (at least with creating beta testers). Anyway, when
this happens, the users connection is also stuck, but clearing it
doesn't allow the user to get back in. The only way we could get the
user to reconnect to the resource was to do an afpstop and afpstrt.
That really isn't acceptable.

Anyone else seen anything like this? Our cluster is 6.5, SP3, with all
post SP3 patches. We may go to SP4a before we go into production, but
I'm still waiting to hear more about SP4a before I make the jump. SP4
(pre-4a) didn't have very good reviews when it came out, so I'm waiting
to hear if all the issues are really fixed.