When I began to work on replication a few years ago, I was convinced that Identity Range Management was only made available in recent SQL Servers for legacy purposes as, prior to SQL Server 2000, the GUID type field did not exist and most primary keys were built on autoincrement fields.
I already saw some of these databases 'at work', and I can say that managing these Identity Ranges in a complex replication topology can be a real PITA, with guaranteed headaches and all-night-long debugging sessions.
I have just read here that Identity Range Management was still proposed as default with SQL Server 2008, followed by this other question on Best Practises and Identity Range Management where #mwolfe02 reaches the conclusion that, in some situations, users cannot insert data in the database unless they are granted db owner's right!
Disturbing, isn't it. So, with the exception of legacy databases, why should someone use Identity Range Management with SQL Servers? In a more generic way, why would someone promote any predetermined\autoincrement identification method in a database?