use case: one mssql instance is tracking the backup age of all it's databases. When the all are backed up at the same moment, they all report their last backup age at the same moment. The data is sent using a batch in zabbix_sender.
The trigger on max backup age of each database fires, ignoring the correlation rule setup for this.
In a test when I separated one database into an earlier zabbix_sender session, immediately followed by the rest of the databases it worked ok. All data was processed sub-second and all reported age's were identical.

I think this is a bug in the handling of trapper data. The first record received should raise the trigger, the rest of the records should be handled by the correlation rule.

I also tried with sending different values for clock in the zabbix_sender file but even up to a one minute difference, the correlation did not kick in. The only way to get this to work was separating one value into a separate zabbix_sender session, which is not usable.