This turned out to be insufficiently useful to adopt for now. We might revisit it later, or pursue it in a slightly different fashion, e.g. we could consider adding some kind of progress information to the status endpoint while in maintenance mode.