Load Shedding Was the Best Distributed Systems Course I Never Paid For
Load Shedding Was the Best Distributed Systems Course I Never Paid For
In March 2026, South Africa quietly passed 300 consecutive days without load shedding. By late April the streak was past 340. This is genuinely good news, the kind we waited the better part of a decade for, and the national mood has been muted and deeply suspicious, like a dog that has been promised a walk too many times before.
Because here is the thing nobody will say out loud while the lights are on. We do not trust it. The inverter is still in the cupboard, fully charged. The gas bottle is still full. We learned something during the dark years, and what we learned turned out to be a surprisingly good engineering education.
A short, painful history
Load shedding started at the end of 2007 and then settled in as a way of life. The worst years were 2022 and 2023, when stage 6 stopped being an emergency and became a Tuesday. Then it turned around fast. Through the whole of 2025 the country recorded only about 26 hours of load shedding, all of it in a brief wobble in April and May. Since then, mostly silence, in the good way.
What fixed it was not magic. It was a long, boring slog. The Generation Recovery Plan grinding away since 2023, big units like Medupi and Koeberg coming back, a real explosion of private solar and wind, and the Energy Availability Factor climbing from the high 50s to around 69 percent. Boring works. Hold that thought, it comes back later.
The accidental curriculum
If you wrote software in South Africa between 2008 and 2024, Eskom enrolled you in a distributed systems course and never sent the invoice. You did not read the Google site reliability book for fun. Eskom set it as homework.
Think about what load shedding actually is from an engineer's point of view. It is a dependency you rely on completely, that fails on a schedule you do not control, sometimes with no warning, for a duration you cannot predict. That is not a power problem. That is every hard problem in computing, wearing a high-visibility vest.
So we adapted, and the adaptations map almost one to one onto real engineering principles.
Redundancy is not optional
The inverter under the desk is a failover system. The little UPS keeping the router alive is high availability for your home network. Nobody in Cape Town calls it that, but that is exactly what it is. You identified a single point of failure, the grid, and you put a redundant power path behind it. Every serious system does the same thing with its database, its servers, and its internet line. We just learned it with batteries first.
Monitoring beats hoping
EskomSePush was a pager for the power grid. You did not wait to be surprised by the dark. You got an alert, you knew the window, and you planned around it. That instinct, watch the thing that can hurt you and get told before it lands, is the entire discipline of monitoring and alerting. South Africans built it into their daily lives before they built it into their apps.
Graceful degradation
When the power went, the good setups did not crash. They degraded. Work moved to the laptop battery, the phone became the hotspot, the non-urgent tasks waited. In software this is the art of failing softly. When a service goes down, the app should lose a feature, not fall over. People here understood graceful degradation in their bones, because they lived it four hours a day.
Assume it will go down
The most valuable habit of all is the mindset. You stopped assuming the power would stay on, so you saved your work obsessively, designed around interruption, and never trusted a long-running process not to get murdered halfway through. That is precisely the assumption that produces resilient software. The network will drop, the server will restart, the job will get killed. Build like the failure is coming, because it is.
Why the inverter stays in the cupboard
Now the uncomfortable part, and the reason this is not a victory lap. The streak is real, but it is not a guarantee. Eskom's own Winter Outlook for 2026 has a base case of no load shedding and a high-risk scenario, officially on the table between mid-May and mid-August, where stages 2 to 6 return if unplanned breakdowns push past 16,000MW. The grid is healthier. It is not bulletproof.
This is the most important lesson of the lot, and the one people forget the moment things are calm. Resilience you build during the good times is the only resilience that helps you in the bad times. The inverter is useless if you buy it during stage 6. The backup is worthless if you set it up after the outage. You prepare while the lights are on, precisely because they are on.
So the gas bottle stays full. The inverter stays charged. Not because we are pessimists, but because we did the course, passed the exam, and have no intention of retaking it.
If you want to keep score
For the days the grid does wobble, our load shedding checker pulls the live stage and your area's schedule, and the Eskom tariff calculator helps you work out what those kilowatt-hours actually cost once the power is back. Use them the way you would use any monitoring tool. Not in a panic, but so you are never caught off guard.
The bottom line
South Africa spent fifteen years learning resilience the hard way, and a whole generation of developers came out the other side with an instinct for building things that survive their own infrastructure. The lights are mostly on now. Keep the habits anyway. The grid got better by doing the boring, redundant, unglamorous work consistently, which, as it happens, is also how you build software that does not fall over. Eskom would love for everyone to forget the lesson. Do not let it.