When a task fails for a reason you have since fixed, you do not need a new run: clear the task instance. Clearing resets it to no status, gives it its retries back, deletes its XComs, and the scheduler runs it again inside the same run, with the same logical date. In the Grid, click the square and Clear task; tick Downstream to rerun everything after it too, and Only failed to leave successful tasks alone. From the terminal: airflow tasks clear <dag_id> -t <regex> --only-failed --downstream -y.
Clearing a whole run (Clear run) reruns all of it. A cleared run of a paused Dag waits in Queued.
Mark as success sets a state without running anything. Use it rarely: when the work was done by hand, for example. Downstream tasks that already failed as upstream_failed do not rerun on their own; clear them.
With depends_on_past=True, a task waits until the same task in the previous run succeeded, so one failed day holds up the next until you clear it.
Rerun the failed busy_stores and everything after it, in every run where it failed.