The load management (including "loss of load") of any modern and properly invested grid is actually rather good,
You are mistaking load management and on/off sequences specifically designed to "break" the grid.
Each grid has them and there are ways to model them - you can get the results either by solving the original optimal control problem for the grid (with different conditions) or from a Monte-Carlo simulation. You can do similar sh*t with most large networks which have active traffic management by the way regardless of what they carry - leccy, packets, gas or oil (*).
It is one of those things which make the difference between a scripting k11d10t and a nation state.
A nation state can employ mathematician(s) which specialize in one of the areas relevant to this (differential games/game theory, optimal control) and give him the relevant data on the enemy so they can come up with the appropriate destructive sequence.
Knowing personally the teachers of the people who will be doing the math on the other side I am giving Beryllia's grid under 5% chance to survive (if the Russians decide to play hardball with it). By the way - we can do it to. Or just ask the Israelis to do it for us.
but the idea of a grid keeling over for days and weeks is mostly alarmism.
That would have been correct if the Smart Metering priests did not throw out a key security criteria in the comms part of that infrastructure during the security review - namely speed of update/speed of reset. As noted by me at the time (and many others) there is a key issue with the idiotic design driven through by the Retail Energy Association and forced upon all of us. Namely, it will take several days to perform a pan-UK switch on/off command update and MONTHS to perform a pan-UK software update. So if you plan your attack properly and backdoor enough SMs to run an attack sequence the grid WILL BE GOING UP AND DOWN LIKE THE BRIDE'S NIGHTY. Regardless of all the advances in automation.
(*)(I have done some of the "tradesman" level coding for the math in this problem space a while back. It is quite an interesting area and most people who think that "realtime engineering load control" without the math can work here do not know what they are talking about as well as what they can face in a real attack.