Something that might help: It isn't one axle twisting. Any one source or sink on the grid can model themselves as a single shaft coupled to the rest of the gri

Something that might help: It isn't one axle twisting. Any one source or sink on the grid can model themselves as a single shaft coupled to the rest of the grid, but for conceptualizing it as a whole you really do have to think of it as a system of coupled machines. One thing that might help with the stability intuition: The generators themselves are synchronous machines, but they have parasitic induction machines deliberately installed in the form of damper bars. Those damper bars mean that the vibrating mode between the generator rotor and the stator field is well-damped. Similarly, the vast majority of total load is in the form of induction motors which naturally have a damped response between the motor rotor and stator field. So resonating patterns and shocks are quite difficult to set up. Even sharp step inputs are attenuated to be not-sharp over short distances. In the US, one common test for grid step response is dropping an entire nuclear power plant. Not one reactor: the entire site. So several GW of electric supply is dropped instantaneously. The limit typically isn't any kind of oscillation, its the ability of primary frequency reserve to pick up the slack before you start triggering under-frequency trips.