ADI Budget Import Process and Performance

We're trying to improve the performance of ADI when importing budget data and are looking for some more insight into how the ADI process works.

It looks like counts are done of the number of records in gl_budget_interface before any inserts take place.  When no records exist in the interface table, the import process seems to work pretty quickly.

However, if records exist in the interface table when an import is submitted, then it seems to do a count of records prior to each insert.  This count prior to each insert did not seem to occur if the interface table was empty when the process started.  The counts prior to each insert cause the entire process to take much longer to complete.