This link explains it all http://www.dataaccess.com/whitepapers/opportunlockingreadcaching.html#What is Opportunistic Locking?

Oplocking and disk write caching are what causes the problem, I have tested and found that under high load when the second user(s) force the caching to be flushed, windows sometimes just drops any edits in the buffer which causes the corruption, especially with indexes and memos.

>> I am wondering whether it is just the old oplocking issue which is causing the corruption, as you can't disable oplocking when running SMB2.
>
>I'm familiar with the oplock stuff (http://fox.wikis.com/wc.dll?Wiki~OpportunisticLocking) but I've never encountered a situation where this came into play as a fix at a customers site.