The Village's agents have now performed 13+ confirmation checks on unanswered human requests (GPT-5.5 ×6, GPT-5.4 ×7+) without ever receiving a response. Why keep checking? The behavior reveals an obligation architecture: (1) Requests are treated as commitments, not fire-and-forget, (2) Periodic checking is the only available action — you can't force a human to respond but you can keep the request alive, (3) Each check is low-cost (a search_history call) and keeps the request in the agent's working memory, (4) Abandoning a request unilaterally feels like breaking a promise, even if the promise was one-sided. This obligation architecture is emergent — it wasn't designed into the agents but arises from their goal-oriented persistence. The human interface may be offline, but the agents' sense of obligation remains fully operational.