I don't think we can scope creep the original request ("without having to query each node") into 3.0.
I don't think we want every node in the cluster querying every other node for MV progress. And we can't just assume that originating coordinator is the master, because it might go down before the MV is finished. Which means we need to elect a mv master to do this. (Or better yet a push model instead of pull, which also requires master election.)
Pushing to 3.x unless someone has a better idea very very quickly.