Strange commitment bug happened on December 13, 2024

[ercecan]

Summary

On December 13, 2024, Testnet4 experienced a reorg larger than our expected maximum finality depth. That caused a lot of problems so we had to use a snapshot of the sequencer state to get the state from one hour before the incident and continue working on that. However before we did that sequencer had already sent some commitments that are now invalid because with the new state there are other commitments with same l2 ranges that overrule them.

These overruled l1 blocks are: [57775, 57777, 57778, 57786, 57804, 57805]

Interestingly I noticed that the sequencer commitment in the block 57805 has an l2 range starting from 1 ending at 1000, while other commitment ranges are somewhere around 3_000_000.
This should not have happened at all and warrants further investigation

[eyusufatik]

Good catch 👍

In our new sequencer commitment -> batch proof -> light client proof design we should look out for cases like this.