@Mikerah, regarding the use of On-Chain Non-Interactive Data Availability Proofs in place of the Master/Cache Shard:
There is first, second and third tier data. First tier is entirely contained in layer 1, highly used, small size public data (even user-created). Second tier is highly used, big size (per item). Third tier is less used data, regardless of size. Second and third tier data can have a seed kept as first tier data and the rest can be stored in Swarm/IPFS.
First tier needs to be available in the VM in full -> a Master Shard (global scope) can do this.
Your proposal is very good for 2nd and 3rd tier.