The hack for me that seems to work is not mapping the rowversion
column for bulk insert/update operations. They happen overnight in a
different process. And leave it mapped for UI where it's mostly needed
because more than on user access the the app via UI. Also imports must
have the final say, and users should not override imported data with
stale data from the UI.
I am not sure about using int, how that works if you write from
different processes (using NH) to same db?