(OTS) Sequencer Heuristic Does Not Split Blocks

[l-monninger]

Summary

The sequencer heuristics try to generate optimal blocks by performing different heuristic passes on the passed-in mempool blocks. However, the final blob bytes that are submitted to Celestia are only generated at the start and are never changed based on the heuristics. For example, if B1[tx1, tx2] is the original block and ends up being split, the two new WrappedBlocks will still store the same Celestia blob bytes containing both transactions. This seems incorrect. Here’s the relevant code:

• WrappedBlock created here with signed blob of the original block
• WrappedBlock’s Split implementation here simply clones the underlying blob in the two new WrappedBlocks
• Finally, the cloned blob is submitted to Celestia here (called from here)

[l-monninger]

I believe the blob field is not actually used further up in the submission. But, regardless this should be changed to construct with try_new or similar. If blob is used, @mzabaluev this could be a candidate for Suzuka state freeze tbh as we might not have triggered the split path for a long time and then suddenly run into this faulty logic.

[l-monninger]

@khokho