What do you mean @filippo_carletti?
In order for normal verify connection we need at least 15 ping probes, and ping somth like Google or OpenDNS servers.
And we shoud have different ping interval for different connections , somth like 1 ISP every 60sec 2ISP every 70 sec , for preventing situation that if two ping starts at the same time one or few packets could lost.
For increase of LSM reaction QA should test specific attributes that control the decision process and time-to-decide :
“max_packet_loss, max_successive_pkts_lost, min_packet_loss, min_successive_pkts_rcvd, interval_ms, timeout_ms”.
And our NS script should do ip ro flush cache on every LSM event , but maybe when shorewall reenable interface they clean route cache.