This feature was something we needed very, very, very, very much.
We manages variety of servers and they are divided into completely different countries or regions.
Therefore scale-up of the poller server was no longer able to handle issues such as server load and network delay, and it needed a way to separate poller.
We could get a good idea about how to separate poller at the below URL.
(http://blog.best-practice.se/2017/10/scaling-observium-horizontally.html)
But what the URL introduced was how to divide the poller instance.
We couldn't meet all of our requirements.
Because we wanted to separate poller into different regions.
Of course, it should be considered carefully to run a remote poller as Adam Armstrong suggests.
it's because all the network environment isn't good.
However, we think that we can prepare for loss of data by arrange the relay server in the middle.
Anyway, we were able to solve the problem completely with this patch.
We finally got the results that what we wanted.
I'm very grateful to Fadly Tabrani for providing this patch.
Mike/Landy is currently working on implementing this featureset into the base poller wrapper. Be careful when updating, since this patch will likely break soon.