Thanks, I'll research the fix further.  Maybe it is an answer to my problems.

dbcc dbreindex was used to rebuild the indexes.

I did update statistics right after it happened.  Also checked index fragmentation and defrag what ever was necessary.  But it still happens. 

The auto update statistics is off.

Yes, we did run profiler.  It shows the usual stuff that we see on a normal day too.

Sp_who2 does not indicate anything.  We do see normal connections from the web server.  We did think of the anti-virus so we shut it off for the time being, but...

I think I miscommunicated.  What I really meant was: is it possible to create a true failover of the witness.  In other words if witness1 fails, then the mirror is watched...

Thanks for the tip.  Setting up an alias helped.  However I have one more problem with a witness server, which is sitting on SQL 05 Standard.  I think the problem...

Turns out it was a problem with the controller.

One of the servers is much faster, the one where queries run for about a second.  The file copy on that one takes couple of minutes vs. about 30 min...

Thanks for the tip, but we already checked memory and cpu utilization.  The cpu and memory is hardly used.  It's very idle...  We have SQL server setup to use min....