We've got it figured out and I'm surprised this isn't more common.  The issue was that we moved from 2003 server to 2008 R2 for all the servers and ran into this fun stuff.  http://technet.microsoft.com/en-us/library/ff686200(WS.10).aspx

 

The saves worked fine, but every once in a while the upload everything event would fire on a different web server event and our reads would get a stale list.  The work around we have now is setting everything to smb1.0.  We tried adjusting all the caches to 0, but there were still random errors.  

 

Now that we have everything set to 1.0 we are running into performance problems with our database backups since they go to the same server.  We are working on tweaking everything to get the right mix of protocols to get everything running.  I'm beginning to think if we could eliminate the disk cache from the uploads and just stream directly to the database it would simplify our setup.