We are on the CDM5.3 and I just recently started supporting the ETL and was not part of the original implementation team. We are implementing in Azure.
For the most part, the ETL is a kill and fill process in the staging tables; and then definitely for all CDM schema tables. Additionally, the primary keys of most of the clinical tables use the INDENTITY(1,1) property, so all keys are re-assigned with each load. Some of the largest tables take between 8 and 60 hours to load.
And now, the team is considering using an incremental load. I was told by the members of the original implementation team that the design intension is for the old database to remain available so researchers can save the cohorts. I wrote a query on the COHORT and COHORT_DEFINITION table and both are empty!
I hope an OMOP expert can provide feedback to my comment above as well as answer if most OMOP implementations use Kill and Fill and if so are they re-setting the primary key fields each time? Thanks in advance!