In discussing the classical capacity of quantum channels, as e.g. mentioned in Wilde's book (see section 20.6), it is possible that using entanglement at the encoder stage can improve transmission rate. In this regards, Wilde discusses a construction given in (Hastings 2009). This seems to rely on considering random mixed-unitary channels, and showing that the corresponding minimum output entropy is non-additive.
Are there other known examples of channels where entanglement at the encoder improves communication rates? Or any sort of more general understanding of some property the channel must have for this to be the case? I'd appreciate both source and explicit examples, if sufficiently simple ones are known.