Perhaps you could expound further on this hypothesis?
I'd always assumed people who spent 3+ years studying electrical engineering had solved this problem. Certainly in Australia (~240V / 50Hz) we don't seem to have a problem with all our clocks being 20% wrong all the time.
The frequency changes are pretty small in normal operation, but on a clock that uses the frequency to keep time they accumulate. They only work reliably because power companies know about them and occasionally deliberately run a bit over or under capacity to make the average match again.
The electrical grid is a bunch of heavy spinning motor-generators that are electrically connected to heavy spinning motor-generators and other loads like lightbulbs. The motor-generators are electrically identical, except that we expect to add energy to one side and extract energy on the other*.
So what happens if the energy added by power plants is less than the energy extracted by lightbulbs and the loads on the motor-generators? Conservation of energy means that we must get the energy by slowing down the generators, extracting their kinetic energy. That lowers the grid frequency.
The same thing can happen in reverse to increase the grid frequency. Too much power generation must increase the kinetic energy of the motor-generators.
* Many of the loads on the grid are intentional or unintentional flywheels, so they may actually add energy to the grid if the grid is slowing, increasing stability.
The only thing that matters is that a clock that expects a certain frequency gets that frequency and not 1% more or 1% less.
It doesn't really matter on a second-to-second timescale how accurate grid frequency is. If you can keep the average frequency right, all your clocks will speed up and slow down in sync, and average out to 24hours per day
But now? It's pretty much just an implementation detail.