Stewart and Sahani showed in 2026 how to sample a discrete law with spikes. A silent unit fires at a rate set by its field, its spike stays live for a fixed time, m, and the pattern of live spikes is the sample. Give each spike an exponential lifetime of the same mean and it becomes a birth-death chain with the same law. The abstract reports that on sixty-three targets the spiking sampler "always outperforms these birth-death processes", and says the sampler exhibits "a discrete form of momentum". Lesson 11 found that a wait with no memory costs nearly twice a fixed one when a fabric chooses which site to visit. A spike's lifetime is the same kind of wait, and one unit shows exactly what its spread costs. Alternating between an exponential wait and a live spike, the unit is a renewal process, and the renewal-reward theorem gives the variance of its live fraction: each phase's variance, weighted by the square of the other phase's mean, over the mean cycle cubed. An exponential lifetime adds exactly as much variance as the wait does and a fixed one adds none, so the birth-death chain needs twice as long for the same accuracy, at every field. A fixed spike ends exactly one lifetime after it began, so its autocorrelation swings negative within a lifetime and the area under the curve halves. That negative lobe is the momentum. Two things change it. A product of two spins squares the lobe, so the energy of three uncoupled units keeps a factor of only one point two. Coupling moves each observable its own way: on a frustrated triangle at inverse temperature one the energy gets most of its factor back, between one point eight and one point nine, and the magnetisation falls to about one point six. The fixed refractory period is the momentum's source, and for one unit any spread in the lifetime spends part of it.