Extending Auctions Without Imposing a Hard End Time

This article proposes a "moving bound" auction design where the effective end time is derived from the highest bidder's Cashu timelock rather than a seller-imposed deadline, enabling auctions that run as long as genuine competition continues. The proposal addresses sniping, free-extension attacks, and unbounded exposure — but reveals two hard problems: self-rebidding is capital-inefficient because each rebid must lock the full new amount (delta-based rebids allow trolls to win with a fraction committed), and exponential bid increments must be anchored to the starting price rather than the previous bid to prevent whales from pricing out competitors with a large opening bid. The biggest open challenge is refund latency: losing bidders must wait for the full settlement grace period to expire even if the seller claims immediately, and we are exploring whether a collaborative close mechanism (analogous to Lightning) could enable early refunds. We are seeking input on increment formula design, collaborative refund protocols, and whether Spilman channel techniques for Cashu timelocks could apply to auction settlement.