Current versions of HA handle this by shutting down the secondary system. I have proved this in testing and actually by having this occur a while back.
Once the Secondary system thought it lost connections with the primary, it started to run its take over, due to a minor inturruption with the heartbeat, once the heart beat comminucations were back, the secondary system shut down.
You then depending on your profile settings for HA, have to bring up the seondary system, and re-start the cluster services on it.