ARD agent keeps relaunching

Question

As discussed here: http://discussions.apple.com/thread.jspa?messageID=7866406
ARDAgent keeps crashing with errors similar to these (I don't have access to that specific machine log at the moment, but this sample is similar to what i saw in the console of the affected machine)
~~~~~~~~~
THIS IS NOT MY LOG -
it is copied from link above - intended to be illustrative of the issue I am having - will post my log when possible - it is quite similar to this one but I can not give specific differences as the machine is not avaialble at this time
~~~~~~~~~
"28/07/08 3:14:33 PM ARDAgent 2613 ******ARDAgent Ready******
28/07/08 3:14:33 PM com.apple.RemoteDesktop.agent2613 LOG: database system was shut down at 2008-07-28 15:14:25 WST
28/07/08 3:14:33 PM com.apple.RemoteDesktop.agent2613 LOG: checkpoint record is at 0/7C38D4
28/07/08 3:14:33 PM com.apple.RemoteDesktop.agent2613 LOG: redo record is at 0/7C38D4; undo record is at 0/0; shutdown TRUE
28/07/08 3:14:33 PM com.apple.RemoteDesktop.agent2613 LOG: next transaction id: 480; next oid: 16886
28/07/08 3:14:33 PM com.apple.RemoteDesktop.agent2613 LOG: database system is ready
28/07/08 3:14:36 PM ARDAgent 2613 Exiting because bind error is not EADDRINUSE.
28/07/08 3:14:36 PM com.apple.launchd178 (com.apple.RemoteDesktop.agent2613) Stray process with PGID equal to this dead job: PID 2619 PPID 2617 rmdb
28/07/08 3:14:36 PM com.apple.launchd178 (com.apple.RemoteDesktop.agent2613) Stray process with PGID equal to this dead job: PID 2617 PPID 2615 rmdb
28/07/08 3:14:36 PM com.apple.launchd178 (com.apple.RemoteDesktop.agent2613) Stray process with PGID equal to this dead job: PID 2615 PPID 1 rmdb
28/07/08 3:14:36 PM com.apple.launchd178 (com.apple.RemoteDesktop.agent2613) Stray process with PGID equal to this dead job: PID 2614 PPID 1 AppleVNCServer
28/07/08 3:14:36 PM com.apple.launchd178 (com.apple.RemoteDesktop.agent) Throttling respawn: Will start in 7 seconds
28/07/08 3:14:43 PM ARDAgent 2623 ******ARDAgent Launched******
28/07/08 3:14:43 PM mDNSResponder31 Client application registered 2 identical instances of service sctxsvr01.net-assistant.udp.local. port 3283.
28/07/08 3:14:43 PM ARDAgent 2623 TCP_ReadThread: bind failed with errno 48 and error: Address already in use
28/07/08 3:14:43 PM ARDAgent 2623 ******ARDAgent Ready******
...And this keeps just going on round and round ....."
The ARDAgent process launches shortly after login. Spirals to about 90% processor. Fails. Re-spawns. Repeat. Watched via TOP on an SSH link and saw it does not matter if there is an active ARD/VNC session. As suggested in the link above, running ARD specific tasks will tend to fail (unless they are QUICK) but the VNC session will run - very choppy but better than no GUI access.
This malfunction slows the machine to a crawl eventually, making it hard to use for basic use of the MS Office Suite, Meeting Maker and Safari. (The machine has 40gB+ free space and 2GB ram. 1.7Ghz Core2Duo).
The Window in my ARD session management gets broken - so regardless of choosing screen 1 or 2, i'll see both but at maybe 70% the "Both" size if I'd just left the setting to show both screens. Once the pulldown is toggled the remoe screen display is mangled as noted for the rest of the session.
Deleted the malfunctioning account temp files in var/folders and renamed the library folder. Upon login I got a delayed login (presumably while the dock and other processes put their temp files back) and a fresh library folder. But no fix on the problem.
the second account on the machine is not affected by these issues. ARDagent works normally if logged out of the "bad" account and into the "good" one.
This user was Migration Assistant moved from 10.5.6 MacBook air to 10.5.6 MacBook. Only the user, not ADMIN account was migrated. Problems did not manifest immediately. Hard to say how long it took as this user does not promptly report problems. I last logged in about 30 days ago and all was fine.
The issue correlates with TimeMachine failing to backup any longer several weeks ago. If you get enough activity on the machine when this process is misbehaving it basically freezes and requires a force reboot (not sure if shell access was still available when that occured as was away from the office and supporting via telephone). Not sure if time machine is now fixed but after moving the users data into the new account spotlight indexing completed normally and all looks stable in top/Activity Monitor.
No new software appears to have been added to the machine between it working and not OUTSIDE OF standard Auto Updates to pre-existing software - Any Apple updates (besides 10.5.7 which is not installed) between March 1 and May 17 were run. As well as MS Office 2004 and 2008 updates. Camino updates. Firefox updates. The User Login items for the malfunctioning account only show Microsoft Office items.
Cocktail was used to clear all system caches. Remove "malicious" files. And rebuild permissions. And to scan for 'corrupt plist files' to no avail.
In any case - seems like there is some simple solution - sure seems like a bad user specific plist file or cache somewhere to remove. Abandoning the user account and recreating it is a painful fix...esp. when the machine is 3 time zones away!
Any insight for a less dramatic and time consuming fix is apprecated. Will post my console report when possible.

Answer 1

AFAIK, ARD's buried inside /System/Library/CoreServices/RemoteManagement/ and you could try removing it. Doesn't autolaunch here, but then I'm not on a network.