>In my experience, lose data could happen in one workstation and one "server" environment. And the data lost not only the last "FLUSH" record but may be 2-30 records.
>If FLUSH not really FLUSH data, why should we FLUSH?

I guess we have different experiences. I can FLUSH on a client machine and watch the drive light on the server operate in step.

My multi-user app with 20 to 25 users, 150 tables, and NT Server has never had a report of "lost records" or a corrupted index in 8 years. The app itself has "cloning" features that will insert 1000+ records at a time when creating work orders, purchase orders, requisitions, etc. Lost records would have had a major impact.