I have to OSS staged in the lab, each referred to by IP by the clients. I find that if I down the primary server it can take over a minute for the next scan request to timeout. I'm seeing 1 minute 15 seconds or so when maximum scan time is set to the default of 45 seconds. When I had only the single server I'd notice it would take quite a while before the pop-up noting the MOVE service was disabled, too.
I don't see anything in the command reference aside from the ConnTimeout value on the client, which oddly says is default to 0 but on my 2.6 deployment is 5 (and won't accept any lower number).
How frequent is the heartbeat? How many heartbeat failures before the client fails over? Can these values be tuned? If there's a failover requirement detected after a scan has begun will the scan be aborted and resubmitted to the secondary server?