Currently, the implementation will perform the maximum jump possible, which is the most efficient for total cycles, but the total latency will be slower once the entire batch can be proven.
Would be more ideal in larger cases to chunk it, such that earlier chunks can be proven as they are available.
Although this might not be meaninful in the case where light client blocks with more signatures and distributed stake, as there will be much more relative signature validation. Benchmarks on Celestia mainnet would likely make this decision (not a breaking change from contract, so either could be picked)
Currently, the implementation will perform the maximum jump possible, which is the most efficient for total cycles, but the total latency will be slower once the entire batch can be proven.
Would be more ideal in larger cases to chunk it, such that earlier chunks can be proven as they are available.
Although this might not be meaninful in the case where light client blocks with more signatures and distributed stake, as there will be much more relative signature validation. Benchmarks on Celestia mainnet would likely make this decision (not a breaking change from contract, so either could be picked)