Thank you @SergioSchena @sumitPanda 
Please refer to case of near real-time (say 15 min), append only configuration of synapse link with Databricks volume being used for raw to bronze ingestion. https://learn.microsoft.com/en-us/power-apps/maker/data-platform/azure-synapse-incremental-updates
1. Shall we rely on timestamped folder's model.json or the latest folder's model.json (excluding the changelog one) to ingest the data?
2. As discussed above, latest model.json can handle historical folders for changing columns but not for changing datatypes, in near real-time how one can unlink/relink and stabilize the ingestion process? 
2. Also, using the timestamped folder's model.json sometimes shows AthenaInitialSyncState  staus as " InProgress" for some entitities and sometimes, inside the timestamped folder, csv data is not matching number of columns with its model.json. What is the recommendation? Is it a known bug?