I'd love to know how it handles adding new shards. Migrating redis data from one shard to another has been a major pain point for us in the past, and has caused us to move away from redis for any big data needs we have where growth can't accurately be estimated.
Yes, but even serializing data at the key level in redis to do it yourself is painful. With 2.6.0, redis provides dump[1] and restore[2] methods which will go a long way to alleviate this, but that version is still in pre-release status at the moment, and was not available at all last year when we had our issues.
