Steps to reproduce:

1. setup a separate cluster from zabbix 6.0.4 running 6.4.5
   identical to first, but running 6.4.5 instead of 6.0.4
   ie: each proxy monitors same hosts in corresponding environment

Result:

The proxies are running 100% http poller on 6.4.5 where they are running

See screenshot...

I've attached screenshots, I'm not seeing anything abnormal in the logs, the proxies are located in the same subnets, but there is a much higher time spent doing the http polling in 6.45 then in 6.0.4 

Expected:
minimal changes to zabbix proxy configuration.  As things are now, I'll have to increase the number of http pollers from 10, which in 6.0.4 was satisfactory.   Please let me know what I can provide to assist in debugging.