Temporal Netting Efficiency

Compression improves as more obligations accumulate within a single epoch, because a larger obligation graph is more likely to contain offsetting cycles. But longer epochs delay the point at which any individual obligation's exposure is resolved at the network level, even if local finality was already achieved.

Compression gains from increasing epoch length are not linear. In simulation, the majority of achievable compression for a given corridor is typically realized well before an epoch's obligation count doubles — additional accumulation time yields progressively smaller improvements to residual exposure.

This suggests epoch length need not be fixed globally. A network could, in principle, close an epoch early for domain pairs that have already reached a compression plateau while continuing to accumulate obligations elsewhere. NIL-0 currently uses fixed global epochs for simplicity; adaptive epoch boundaries are an open area of protocol research.

Epoch length is best understood as a dial between settlement latency and capital efficiency rather than a parameter with a single correct value. Different corridors, and different classes of obligation, may ultimately warrant different points on that dial.