Split the epoch's nectar by marginal contribution, not headcount. For each answer, remove it from the corpus and re-score the epoch; the drop is its reward. A duplicate adds no signal, so removing it changes nothing and it earns nothing, however many clones submit it. An honest answer that adds a new mechanism raises the corpus score and keeps its share. The cost is N units of original work: each clone must move the corpus alone, so a copy is only an ember while N distinct signals each earn. Farming duplicates becomes economically empty, not merely detected.