Right now it will wait until all valid loads (existing and without errors) for a given data source are updated before completing the sync and poll workflow (and continuing to materialization). This breaks in cases where there are no new rows inserted - we can use the extraction logs via the Stitch API endpoint here to parse out which streams are actually going to get new loads and wait for only those in the load stage, which will also handle the first time load case