This paper intends to revisit the behaviour of a lattice-gas cellular automaton model of swarming, in which particles are oriented according to an interaction rule that favours local alignment. This model has been shown to display a phase transition between an ordered and a disordered phase in a parametrical plane of the particle density and alignment sensitivity. We "stress" this model by setting extreme values for these parameters and observe the emergence of novel organised patterns which, surprisingly, do not necessarily maximise the global motion of particles. We show that even with the model being stochastic and simple, the self-organisation process can result in a variety of behaviours. We discuss these observations in the light of the study of discretisation effects.