The PTLC one sounds vaguely familiar. I think it had to make a ton of tradeoffs to limit an intractable number of pre-signed txs.
The paper I started reading was using timouts and every LN transaction was actually creating a valid Bitcoin transaction with a shorter timeout with the theory being that the shortest timeout transaction would be mined on-chain before any of the older states.
Oh interesting. I hadn't thought about doing more with time-iterated games. Time is one of few trust-minimized levers we have for this stuff though.