We are currently looking at deployment scenarios for ArcGIS for Server. We are finding that having local copies of file geodatabases on GIS servers is ideal for building map caches and performance. ArcGIS Help documentation recommends using geodatabase replication for publishing local copies of data to GIS servers (especially when in a new projection (e.g., Web Mercator Auxiliary Sphere). In our production environment we have an editing ArcSDE geodatabase and a publish (read-only) ArcSDE database that receives updates via python scripts that first truncate the destination table in the publish geodatabase and then append features from the edit geodatabase to the truncated table. We follow this method to avoid deleting feature class metadata using the create/delete method of updating features.
I have a concern about using one-way replication in this scenario for local copies of file geodatabases on GIS servers. How will we be able to manage the GlobalID column when we are truncating and appending? The GlobalIDs will be different between the data source and data target of the one-way replica. What are some potential workarounds to this scenario?