essentially, the X scraper fails, then it continues with the rest of the steps (including creating a doc)... then, after the entire workflow runs, the agent tries to re-run the X scraper which then re-runs those same downstream tasks, creating another doc unexpectedly