>Yes, I've considered it, but I know this is not the case. Subsequent
>threads running the same query post their data just fine. Even though
>it is slow, because a previous thread has caused the deadlock problem

Sorry to keep hammering on about this, but that sounds like garbage collection. The thread that triggers the GC process is kept busy and will not return results until the process is done, any subsequent threads will run just fine, although obviously slower due to the server workload of the GC thread.
Could you let us know what your sweep interval is set to and post your header page stats.

Regards
Gary


[Non-text portions of this message have been removed]