I'm curious why they chose CollectD and not Prometheus or another pull-based ("scraping") solution. Pulling is more compositional than pushing and requires fewer machines to be touched when monitoring configuration is changed.
Probably due to the size of deployment. I'd go to Prometheus in a larger/dynamic environment for flexibility, but I'm happy using telegraf on a home network where I configure things by hand when I'm logged into each pet machine.
The author responded in another thread, it was just due to being familiar with CollectD already