Makes sense, thanks AJ for the explanation.
I wonder however if it is worthwhile to even consider cases in which an attacker can keep things published but hidden from a specific victim, as that is quite the rabbithole, and I’m not sure it is that easy to pull off.
As you mention yourself the threat of a watchtower should be enough to keep the chances of succeeding to a minumum. Notice that if the hidden TX is discovered before being mined by the victim, then they can queue a reaction for the same block, forcing the attacker to add more funds to retry with a new commitment, hence why I talk about game theoretic threats despite there not being a direct penalty for attempting to cheat.
The added complexity for penalizing and/or forcing a sequence may outweigh the benefit.