BD, I wasn't a relay technician, but sure did my share of power management in more localized applications. Power management works the same way with utilities, but it's just done with bigger ones and zeroes and fancier computers. I've done loss-of-source load shedding, demand load shedding (by projecting peak demands), and like you pointed out, had to accommodate every one of those little magic numbers that most customers never see. I even modeled the performance curves on a couple 20 MW generators in a paper mill in southern Mississippi, so I could take them to max limits when the mill was in full production and on a lowered peak demand with a power factor penalty. I took control of the two vent valves so I could vent the low-pressure headers and let the generators put out more. Normally, I'd just cram every ounce of steam I could through the condenser by calling for more flow through the condenser stages to the turbine controls. Costs a lot less to do that than vent it. This is not easy to do when the generators have old analog EHC turbine controls and analog Exciters. A slow response in these systems may as well be no response. When grid systems were down, that papermill would back off on production, and provide power to the local town (where most of the employees lived) until Utility service was restored. Being on the Gulf Coast, you can imagine how often the distribution grid is disrupted by hurricanes.
I put in energy management systems all over the US, some of which had as many as 8 generators and 4 utility ties. That work took me to some interesting places. Most of them were based on the different type of power purchasing contract the customer had. The old peak billing systems are all but gone now, except for large commercial users with no cogeneration, and even those customers typically have the newer LIP contracts that allow them to take what they need until the Utility calls them and says back off.
The only thing I meant by what I said about those boundaries is that the distribution lines do actually cross those boundaries with some pretty simple switching. The only things that stops the power flow is money, and someone in control of that money, or known distribution issues that would compromise the neighboring provider. If ERCOT can do rolling outages (which apparently didn't work very well) they're a lot less likely to purchase from SoCo (SERTP), TVA, or any other commercial zone. They'd rather take the hit. The losses from the outage are a drop in the bucket compared to the price of sustained purchasing from and adjacent zone. It looks better at the end of the quarter when there's lower variable costs in the spreadsheets. I don't remember any reports of widespread outages in Louisiana, and apparently they had quite a storm (for Louisiana) too.
The worst problem in the south is when ice builds up and pulls down the lines. I don't care how much power your neighbor has then, you can't use it if your distribution system is compromised.
