When the proCube server goes into production, data loads must be planned carefully. Since data loads can potentially change the fundamental data from which a cube extracts (as well as sometimes being timely operations), the best practice in production mode would be to schedule data loads in the hours that users are not actively accessing the proCube from V-i-Performance. Rebuilding while users are actively accessing proCube is not recommended, as it can potentially lead to data corruption issues (due to memory over-usage), as well as present application responsiveness issues for users.
Finally, in production mode, since manual hierarchies can be created after data loads, it is recommended that Rebuild Hierarchies in proCube Functions be used only when necessary. Again, from a data load perspective, the only time hierarchical data will be loaded with new data is with a Stage_GL_Rollups file.