At the moment Treeherder UI is fetching the tree status from treestatus.mozilla.org. It would be nice to store that information internally and serve it via the api/socket.io. The would decrease the load on treestatus and would let us handle those repos that are not handled by buildbot. Ideally we will use a etl task to poll that info out of treestatus.
I'm not sure if we still want to do this. treestatus.m.o is quite happily handling the load - it's more does this give us any advantage if Treeherder can track the state? If tying job results with treestatus more closely is something that is definitely useful/we want to do, then perhaps we should fix bug 1075129 instead?