The tokenomics analysis is that as a smart contract gets more valuable, it attracts more of the native currency (eth, xdai, safecoin, whatever). Smart contracts of the type ERC20 or ERC777 might have to additionally send safecoins to indicate the size of the value transfers. (Since we don’t really know how many safecoins really should equal one USDC or DAI token).
After each transaction, as more and more such value ends up in a section, that section will have to recruit MORE vaults to secure it (something along the lines of log N vaults to secure N amount of value).
Also, every time a transaction begins (some user signs it with their key and pays enough safecoin to make it happen) the section hosting the smart contract, and its vaults, may pull in OTHER sections into doing one instance of consensus because the transaction would be atomic across several sections (shards).
Whenever a smart contract hosted by one section calls methods of a smart contract hosted by another section, it would be pulled in. Typically smart contracts do not call outside their web of trust (whitelist) because the called function may do a while(1); and kill the transaction with out-of-gas. So you will have “affinity” between some contracts and the network can learn who to pull in for what method, and even optimize that over time. After the transaction ends, each section checks how much safecoin or ERC20 balances it ended up with, and adds or reduces its recruitment of vaults to log N of that number. It would be a market for vaults to join.
Smart contracts would start out with zero value, but as they attract more safecoin, the section hosting them would attract more vaults. Or alternatively, we can keep section size uniform but sections can recruit other sections to join into a consensus with them, or disband as coins move to another smart contract address.