All data about host availability have been moved from the host level to the level of individual interfaces. During database upgrade all previous host availability data will be lost. New interface availability initially will be set to 'unknown' and then updated after some time.
Note that the server-proxy data exchange protocol has also been changed.
Calculated, aggregated and internal checks are now performed by the new history poller process. The StartHistoryPollers value should be increased if history pollers are too busy, but should be kept low if possible to avoid unnecessary connections to database.
New availability manager process has been introduced. All processes queue host availability updates to the availability manager and that queue is flushed by the availability manager to the database every 5 seconds.
You can monitor the new processes using the zabbix[process,<type>] internal item.