We have experienced that Azure Synapse Link stops working for certain, usually large, tables. In some cases we have used append-to functionality with a datalake to utilize this as an archive for the data. And hence the data in dataverse has been removed.

When talking to support, they recommend that we remove the table and restart it. That would break the archiving functionality we are using it for or any other use that relies on the data to be coherent with the data in dataverse.

Hence an improved failure handling should be implemented where it self-diagnoses errors, tries to restart etc. And if a table is stopped and then restarted again, it should understand this and start up from where it left off.