In order to reduce the amount of wasted compute, we can to use server jobs for waiting for helix jobs to complete, so we don't tie a build machine up just waiting for a response.

I'm not entirely sure of the mechanics here, because server jobs are incredibly limited (I don't think it's possible to, for example, access the source code, or any of the artifacts published by the build).

Related write up here: https://gist.github.com/RikkiGibson/4423b6470881ea4111a807e8db3a47a1