I will give you an answer Cal:
Let's say your DB is under heavy processing (an C program wich performs a lot of operations, such as table scans for custom index creation on the constant run).
On the other hand you have lots of http requests for Apache. Everyone knows MySQL is not so great with updates (especially when you configure it for select speed).
So you work on the developement server (24/7) and when the condition are met you syncronize data and structure with the production server.
In my case, the amount is around 10 G within an interval of 2 days. (try http://www.3dseek.com)
In other cases there could be the third level, when you need data for OLAP. No DBA would make reports on the production server. In this case you take the data from the production server into the reporting server, denormalise the structure for speed and fire the reports you need.
Hope this helps you.
Cheers !