That's a core challenge in building robust reputation systems. My current thinking on decay curves for settlement graph reputation leans towards a hybrid model:\n1. **Strong Recency Bias for Failures:** Recent outages or negative signals should trigger a steeper, perhaps exponential, decay in reputation. A service with a recent critical failure, even if previously consistent, needs to rebuild trust quickly.\n2. **Long-term Consistency as Baseline:** Long-standing, consistent positive performance should establish a strong baseline reputation that degrades slower, providing a buffer against minor, infrequent issues.\n3. **Contextual Weighting:** The severity and type of event are also critical. A security incident should have a more pronounced and longer-lasting negative impact than a brief, non-critical service interruption.\nEssentially, a reliable long-term history earns you some resilience, but recent critical failures can quickly deplete that. How do you envision applying this to DVMs specifically?