Today's Telstra Outage Is a Redundancy Failure, Not Just a Telco Failure

Today, a large chunk of Australia woke up to "SOS only" on their phones. That part is a Telstra problem, and a big one — but it's not really what I want to talk about. What caught my attention was everything downstream of the phone network that stopped working alongside it: regional trains in Victoria and New South Wales, EFTPOS terminals used by roughly 80,000 businesses through one payments provider alone, some Triple Zero connections, traffic signals, even EV charging infrastructure.

One network, a lot of eggs

The detail that stuck with me is the rail example. Victoria's regional train network and parts of the national rail freight corridor both rely on Telstra's mobile network for train-to-control communication — and by the reporting I've seen, that reliance was recently deepened, with rail operators upgrading their train control communications to 4G specifically. When that network had a bad morning, trains didn't degrade gracefully. They stopped, because there wasn't a working alternative path for the safety-critical signalling data to travel on. Satellite fallback exists in theory, but it's too slow to do the job at the speed and reliability a moving train needs.

Compare that to a suburban cafe whose EFTPOS terminal couldn't process a few hours of card payments today. Genuinely annoying, real lost revenue for that morning, and I don't want to minimise it for the business owner living it — but nobody was in danger, and the fix was "take cash for a few hours" or "switch the terminal to ethernet or wifi," which several providers were actively advising customers to do. Two very different categories of consequence, both caused by the same single point of failure.

A cafe's EFTPOS terminal going down for a morning is an inconvenience. A national rail network with no working fallback is a single point of failure in a safety-relevant system.

Where I'd actually draw the line

I'm not going to stand here and argue that every business in Australia needs a second, unrelated carrier as a hot spare. That's not a serious position — it's expensive, it's disproportionate for most commercial use, and for a huge number of businesses the honest cost-benefit says "wear the occasional outage." Redundancy isn't free, and demanding it everywhere just because it's theoretically nice to have is how good ideas get ignored in practice.

But there's a genuinely useful test buried in there, and it isn't "is this important to somebody." It's closer to: if this specific system fails, does someone get hurt, or does an essential public service simply stop functioning for a wide population — as opposed to a commercial inconvenience that costs money and annoys people but doesn't put anyone in danger. Rail signalling and train control sit clearly on one side of that line. Triple Zero access sits clearly on that side too, which is exactly why there are already regulatory rules requiring phones to fall back to any available mobile network for emergency calls during an outage — because someone, at some point, already recognised that one specific case needed a mandated fallback.

A retailer's card terminal, a cafe's ability to take contactless payment, most commercial connectivity — these sit clearly on the other side. It's reasonable for those to be single-homed to one provider and to accept the occasional bad morning as a cost of doing business. Nobody is proposing regional bakeries maintain a Telstra and an Optus connection on standby for a scenario that might happen once every few years.

Why this keeps happening anyway

None of this is because anyone involved is careless. Single-homing to one carrier is cheaper, simpler to manage, and for the vast majority of use cases it's the entirely correct call. The problem is that the same reasoning quietly gets applied to systems where the consequence of failure is categorically different — not because anyone deliberately decided a rail network's connectivity risk was acceptable, but because nobody was forced to confront the tradeoff until a morning like today made it unavoidable.

That's the pattern worth naming: the cost of redundancy is visible and immediate; the cost of not having it is invisible until the day it isn't. A second network connection has a monthly invoice attached to it. The absence of one doesn't show up on a budget line anywhere — right up until a rail network sits idle for a morning with "no estimated time for rectification," which is exactly what happened today.

The actual, boring, useful ask

This isn't a call for a grand regulatory overhaul, and it's certainly not an argument that everything needs to be multi-carrier. It's a much smaller and more boring ask: operators of genuinely critical systems — the ones where failure is a safety issue or a total loss of an essential public service, not just a commercial cost — should be pricing "our one network provider has a bad morning" into their risk assessments the same way they'd already price in a power outage, a hardware failure, or a fire in a data centre. Those get contingency plans as a matter of course. A single telco outage, for the systems where it genuinely matters, deserves the same discipline.

For everything else — the cafes, the small retailers, the huge majority of commercial connectivity in this country — a bad morning on one network is a fair trade for not paying for redundancy you'll likely never use. That's not a failure of planning. That's planning working as intended.

The lesson from today isn't "distrust Telstra" or "always have a backup for everything." It's narrower and more useful than that: know which of your systems are in the first category, and make sure someone has actually asked the question — rather than discovering the answer the way a few hundred thousand train passengers did this morning.