After my Zabbix installation was upgraded from 3.4.14 to 4.0.1 I've noticed that many items of type Zabbix Agent (active) stacked in queue. System time on clients OS and Zabbix Server OS configured to local NTP server (Windows Server 2012R2) and match. But, after a few days I can see growing queue of active checks. If I restart agents on clients, it's become OK, but returns to same state agein in couple of days.
31.10.2018 Zabbix Server was upgraded to 4.0.1:
In zabbix_server.log file I can see many log_client_timediff() messages in debug mode, thougth time are same: