CIDs are just a way to fingerprint a particular sequence of bits. CIDv0 was just the 58btc-encoded SHA-256 hash of these data. CIDv1 will just enable to to pick the hash function and the encoding you like. (IPFS will soon use 32encoding of SHA-256 as a default, but you will be able to use the one you like, really). Basically, CIDv0 was the fingerprint of the right thumb of the data, CIDv1 let you identify the data by any finger.
But it has nothing to do with the datastructure, which is IN the content-addressed bits.
IPFS already transparently splits what you uploads into chunks. I think they are 256KB-long. It is managed under the hood.
For more DAG magic, maybe you can look up IPLD.
IPFS also already checks that received chunks are correct. That’s one of the perks of content addressing.
For streaming videos or playing them while downloading, look at :
… and there is one great video about live streaming to a big audience in the same local network, but I can’t find it again…