i don't work with many zabbix production instances, but as i think and as i see in our environment we have many similar data in DB (and yes we use compression in InnoDB it's cool)
Every ping save new data to DB, every get zabbix version from agent we can continue indefinitely.
(Yes we use zabbix_send and snmp traps as work around for some items, but this is only workaround)

Maybe it can break trigger logic, but we can just save in DB only changes of value.
i.e. i open cpu load, and i don't want see every value in time, i just want see, what in 14.35 value reached 100% and its don't change until 15.15, after value has changed to 14%
And we can store in DB instead of 15.15 - 14.35 = 40 (one check per minute) only 2 values start & stop, and its cool, we can decrease DB size in many times.