I think in both cases (repl and dist) it still may make sense in some cases.  E.g., in dist, if a node joins, existing owners could, rather than push data to the joiner, just push a list of {key: version} tuples, which may be significantly smaller than the values.  The joiner can then load stuff from a cache loader based on key/version - we'd need a new API on the CacheLoader, like load(Set<KeyVersionPair> keys) - this can be implemented pretty efficiently in many cache stores such as JDBC.  The keys that the cache loader doesn't retrieve would need to be pulled back across the network.

Certainly not high prio, but something to think about for Infinispan.next().