Suppose I have an airflow dag file that creates a graph like so...
def get_current_info(filename) current_info = {}
The DAGs are evaluated in every run of the scheduler.
This article describes how the scheduler works and at what stage the DAG files are picked up for evaluation.