Bug#352830: Race condition

["Chris AtLee" <chris@atlee.ca>]

I'd like to note that this problem also manifests as a race condition.
 I have some scripts that execute multiple 'scp' commands in parallel,
and I quite often see some of them fail with "ControlSocket ...
already exists"

It seems like what happens is scp shuts down the connection before
deleting the control socket file, thus preventing other scp commands
from running.  Or maybe the socket file is created before the
connection is active?

I can reproduce pretty easily with this:

#!/bin/bash
for i in `seq 10`; do
   ssh host date&
done