We're having the exact same issue. Currently have to periodically run manual scripts to clean up all the stuck flows.