Fix error handling for initial zone refresh in persistent search.
    
    Old code terminates watcher thread in case of error. Now initial lookup
    is restarted after reconnect_interval seconds.
    
    Signed-off-by: Petr Spacek <pspacek@redhat.com>