We're facing the same issue.. Tried to create a tokenmanager class to force update the refresh.. No luck so far..

              connection_properties["accessToken"] = token_manager.get_access_token()  # Ensure fresh token
              connection_properties["queryTimeout"] = str(timeout)
              
              df = spark.read.jdbc(url=azure_sql_url, table=query, properties=connection_properties)
              return df

Anybody managed to solve this? Databricks runtime 14.3lts , latest msal