Handling the metadata was the really big challenge for our DW. 250 trials, each with 30-40 distinct tables, adding a new trial a week, I had to develop a dynamic ETL tool that was fed by a metadata repository. Finally, the metadata of an individual trial could and would sometimes change in the middle of a trial.

Unfortunately, Kimball was worse than no help at all. His methodologies assume certain things, like static metadata, a small number of source systems, etc.