If a host becomes unavailable by definition, it's items, be they active or passive, should no longer be considered to be in queue.
Switching a whole lot of items from passive to active agents, we noticed that active items from unavailable hosts are held in queue indefinitely. This is more than just misleading, as we'd certainly need to know if passive items are delayed (performance issue), but these ghosts pollute the data to the point of being useless.
The queue, as seen with the web frontend and zabbix[queue] should be cleaned of unavailable items and/or split by item.type.
As a frontend improvement, queue details could be sortable by any field and the hard 500 item limit is an inconvenience.