With a 2.6 system this problems will be simpler to detect, because at this point we have explicit (configurable) limits for clients, of type normal, pubsub, slaves, so the connection is dropped if the amount of pending data goes over a given limit (and the event logged).

This added a bit of complexity to a few code paths that now require to estimate the size of objects, data, and so forth, but I think this will pay back in the long run.

This blog post makes me wonder if we should implement compression in the replication link as a feature in the future, or at least binary encoding. In short there is a plan for a binary version of AOF / replication-link that can save a lot of bandwidth.