f5 failover issue

This is very weird as you say, because normally each BIG-IP in a cluster checks the backends using its own physical Self IP.

So normally, both devices monitors continually all nodes. And when it failover, VS are already in a stable state.

Could you check the health monitor frequency ?