I think some mistakes were made on this topic. There is finality once the operator submits the block to the root chain. Before that there is no finality and a receiver of a coin cannot be certain that coin hasn’t been double spent. Once the Merkle root of the sidechain block has been posted to the root chain, there is economic finality as if anyone tries to cheat the system, anyone can challenge stopping the invalid exit.

This statement from the article also seems to be inaccurate:
“Invalid exit penalty: We can punish the operator for allowing invalid exits to go through, and slash funds in the bond.”

The operator isn’t the one that gets his bond slashed. It is the entity that tried to make an invalid exit.