There's another wrinkle to the basic explanation that should probably be mentioned. The pulsations are cyclical, they are not just a shift to a new equilibrium in which the instability goes away. To be cyclical, there must be a restoring mechanism to go with the unstable mechanism. The restoring mechanism is that once the star expands sufficiently, the mechanism that was allowing it to expand (the absorption of heat by absorbing radiation by bound electrons in metals) eventually short circuits, because once the star expands enough, the density drops and the ability to absorb radiation also drops. So the heat that was absorbed early in the cycle is released late in the cycle. This has the net effect of adding heat when the gas is compressed, and releasing it when the gas is expanded. Sound familiar? That's how a piston engine works! If you add heat when compressed, and release it when expanded, you create a cycle that does a net amount of PdV work. This work will go into pumping up the amplitude of the oscillation, until it reaches some kind of nonlinear saturation when irreversible mechanisms kick into dissipate the work. Often, transporting heat across a temperature difference is the best dissipation mechanism. So when the driving layer of the oscillation is no longer at the same temperature, heat transport within the gas will dissipate the work being input, and the amplitude saturates.