RAC is vastly different for the same machine on different projects. It cannot be used as a measure like this. If you mean that the RAC bar will change depending on the project, how do we define an 'average' system?
Also, what if a few members point huge clusters at a project? A lot of compute would be provided, but the 250 active member requirement will not be reached.
Thank you for your comments @dutch. To futher develop the minimum output I created the above crude measure to provoke the discussion. I'll take it further on github.