distributors should wait a delay to retry the connection to the backend

Each retry produces an entry in the log, a missconfigured service was retrying so frequently that has produced 10GB of logs and fill the disk.