Unstable fluid interfaces that deform into fingers, or break-up into drops, often do so with a remarkably regular pattern. This is true of a variety of interfacial instabilities, such as the capillary break-up of liquid jets, and the thermocapillary driven rupture of a thin film on a hot plate. It was Lord Rayleigh who, in 1879, first hypothesized that the emergent pattern is determined by the wavelength corresponding to the fastest growing linear mode. While this paradigm has worked well in practice, its success is quite surprising, given that nonlinear effects become important well before the interface ruptures. What then is the role of nonlinearity in selecting the pattern? And what would happen if the linear growth curve has more than one peak? These questions will be addressed in this talk, by way of targeted numerical calculations and reduced-order models. The results reveal a counter-example to Rayleigh’s paradigm, and show that interfacial patterns can in fact defy linear growth.