Context. Meteoroid clusters represent a unique opportunity to study processes of meteoroid disruptions in interplanetary space. We follow our previous work about the September epsilon Perseid (SPE) meteoroid cluster from 2016 with a detailed analysis of the observed data and cluster formation conditions. Aims. Our goal is to determine ejection velocities of the cluster members and SPE's age, as well as to estimate the most probable formation process. Methods. We precisely determined mutual positions and masses of all meteoroids including the errors. We assumed that the massdominated meteoroid is the parent body of the cluster and that the observed positions of meteoroids are controlled by the ejection velocities and the action of solar radiation pressure. A formula for the dependence of meteoroid ejection velocities on the mutual positions, masses, and cluster age was derived. It was assumed that the time at which the initial kinetic energy of all meteoroids reached a minimum value corresponds to the age of the cluster. Knowing values and directions of ejection velocities together with meteoroid masses then allowed us to determine the most likely process of cluster formation.Results. The meteoroids occupy a volume of 66 × 67 × 50 km and are shifted in the antisolar direction by 27 km relative to the parent meteoroid. The age of the cluster is 2.28 ± 0.44 days. The ejection velocities range from 0.13 ± 0.05 m s −1 to 0.77 ± 0.34 m s −1 with a mean value of 0.35 m s −1 . The ejection velocity directions are inside the cone with an apex angle of 101 ± 5 • . The axis of the cone is ∼ 45 • away from the solar direction and ∼ 34 • away from the mean direction of the flux of small meteoroids' incident on the parent meteoroid. Formation due to the separation of part of the surface due to very fast rotation is the least likely thing to occur. We estimate the rotation frequency to be about 2 Hz and the corresponding stress is several orders of magnitude lower than the predicted strength limit. It is also difficult to explain the formation of the cluster by an impact of a small meteoroid on the parent body. However, this possibility, although not very likely, cannot be completely ruled out. The most probable process is the exfoliation due to thermal stresses. Their estimated magnitude is sufficient and the derived ejection velocities are consistent with this process of formation.