sql server replication: How does SQL 2005 handle Two-way merge with identity columns?
How are identity columns handled with incremental numbers?
Does the merge process prevent duplication of numbers?

With automatic range management (default in 2005) the publisher and
subscriber are each given mutually exclusive ranges of identity numbers. Some
DBAs in SQL 2000 used to prefer to manually set the identity seed to avoid
clashes (eg odds on the publisher and evens on the subscriber) which then
avoids any potential for a subscriber to run out of a range of numbers -
seems to be set up this way less frequently now.
Cheers,
Paul Ibison