In order to meet changes in demand, the generation authority always has more generators running at any time than is needed to meet the present consumption.
Assume we just have rotating generators. The instant a load is switched onto the grid, the current that flows creates a braking torque in the generators, and they start to slow down. The instantaneous energy needed to power our load comes from the kinetic rotational energy stored in the generators.
As the generators start to slow down, the speed regulators kick in, and increase the mechanical power to the generators to maintain their speed.
The generation authority would notice that supply A was starting to get close to its maximum, so would start up supply B, just in case there was a further increase in demand. If that increase happened before B was up to speed and connected to the grid, then you would have at best a voltage sag (brownout) as A failed to cope, and at worst a power cut.
If a further increase did not happen, then B has been started, not supplied any power, and been a net cost. The margin between the maximum capacity of all the running generators, and the present consumption, is known as the 'spinning reserve'.
This is why forecasting is a very important part of running an electrical supply. You need to be able to guess from the weather, and from the TV schedules, when, and how big, the load demand spikes will be, to have enough, but no more, extra capacity already running to cope. Too much spinning reserve will reduce the profits. With too little spinning reserve, an unexpected demand spike could leave you having to explain an embarrassing and unnecessary power cut.