There are three ways to handle the difference between relatively fixed base load power generation and highly variable grid usage.

First, you can add variable load power generation facilities, that are less efficient overall, but can change output levels more quickly.

Second, you can add an energy storage and return capacity to the grid. For this, you need a reversible process and good efficiency.

Third, you can add variable usage. "Smart grid" appliances, particularly air conditioners, are good for this. But an energy-intensive industrial process co-located with the power plant, that could be scaled up or down automatically, would be better.

The fact that most solutions to the problem of handling variable load when it is most efficient to have near-constant production are of the first variety is easily explained by the economic factors. A power generating company understands power generation better than any other industry, therefore has a bias towards solving its power generation problems with different kinds of power generation.

It is only relatively recently that computing made it possible for automated usage negotiations could take place between producer and consumer on a scale that would be significant. And we still don't quite have batteries good enough to make peak-leveling storage practical or cost-effective. Pumped hydro is about the only game in town, and that depends heavily on appropriate terrain.

every spinning generator has stored energy in the rotor's inertia. Some turbine-generators have flywheels on the shaft to increase the inertia and decrease changes in speed that happen when there is a mismatch in load (power out) and generation (power in). When there is a mismatch in load and generation on the grid every single generator slows down or speeds up. you can monitor this by measuring the frequency of the voltage anywhere on the grid, it will be the identical measurement at all times (except where DC interconnects are involved).

Every generator should have a 5% speed-droop characteristic where a 5% change in speed results in a 100% change in output. So if you were running your hydro plant at 80% output with 60 Hz speed setpoint, and some other generator tripped resulting in more load than generation, and the frequency on the grid dips to 59.9 Hz (a 0.1% change), your hydro plant would increase the output by 1/5% * 0.1% = 2%, to 82%, as would every other generation source that can control its throttle, and once again generation and load are matched, but the frequency is now 59.9. A couple of special power plants (running in Automatic Generation Control (AGC) mode, have control systems with a big picture view of large portions of the grid) will increase their output to restore the lost generation, raising the frequency back to 60.0, and re-balancing whatever power flows were scheduled across the various transmission interties for that hour.

Any source where you can control the throttle should have this 5% droop characteristic. wind turbines and solar distributed generation without batteries can't provide it, so not only do we need more dispatchable resources like gas, hydro, etc to be able to quickly make up the variations in wind and solar output, but the conventional plants either can't run as close to 100% output or have to have increased droop to make up for the sources that have zero droop.

The fact that peakers are more cost efficient than flywheels or other storage tells you how expensive storage is. Plus it tells you that our electricity grids have been very robust to not need storage.

Alternatively it tells you how heavily subsidized fossil fuel based peaker plants are. See the recent IMF report (not a group known for being idealistic hippies) for some numbers on that.