Currently Zabbix proxy is the most tricky component to set-up High availability.
If we have the Proxy pair working in Active/Passive with VIP, it is highly possible that during failovers, failed node will have some unsent history in proxy_history table. When after some time proxy failovers back to this node, it will send this information to the Server, that will lead to false alarms and overall trouble with triggers etc.

Propose to have a configurable parameter in zabbix_proxy.conf that will force to drop this information before the start of a proxy process. Most involved tables - proxy_history and proxy_autoreg_host