The recipient can force close at any time by signing the latest channel state and broadcasting it. They have no incentive to sign an earlier state because the balance on their side of the channel monotonically increases. This is why Spillman channels are unidirectional. This design removes game theory from the payment channel security model.

No justice transaction, no toxic data, no problem. LN-symmetry also removes the game theory from lightning channels but it requires a soft fork and we're all too busy fighting over stupid bullshit to come to consensus on a consensus change. Soft forks are a dead end until someone changes the game and refocuses bitcoiner attention to actual problems that we can solve. 👀

You are correct that the sender cannot unilaterally close the channel before the timeout. After the timeout, they can force close and retrieve the entire balance, so the onus is on the channel recipient to close or renegotiate before time's up.

With good lightning integration, the sender can simply drain the whole Spillman channel balance into a lightning channel. Once the Spillman balance is all on the receiver's side there is no reason to keep the channel open. There is a very good reason to close it early: so the sender can't claw back the whole balance after the channel timeout.