Redundancy is not route diversity: two cables, one weekend

Two links only count as two if they share nothing physical. Over the weekend of 8 and 9 August, two submarine cable systems failed off Perth — according to their owner, almost simultaneously and close to each other. On paper they were separate routes. In the water they ran through the same area, and that is where both of them failed. We come across the small-scale version of that story regularly, in twenty-person offices.

What happened off Perth

SUBCO, which owns both systems, detected shunt faults on Indigo West and Indigo Central, two cables that together connect Singapore and Sydney with a landing at Perth in between. A shunt fault is exactly what it sounds like: the outer insulating layers get damaged and the interior components are exposed to sea water. They are commonly attributed to ships' anchors and fishing activity.

Bevan Slattery, SUBCO founder and co-chief, posted about it on LinkedIn that same weekend with a ship-tracking chart, describing the activity in the area as "suspicious/coincidental". He is the one who told iTnews that the two faults occurred near each other and at a similar time, and the one who was careful not to accuse anyone: "this could be a coincidence or an external aggression event", he wrote. The Australian Federal Police confirmed to iTnews that it "has received a report of crime in relation to this matter" and that "this report is currently being assessed based on the information provided". Neither the company nor the police have attributed this to anyone, and neither will we.

Anyone running a network can leave the question of who did it to the police and keep the technical fact: two systems bought as two routes behaved like one.

A protection zone does not reroute packets

Both cables were laid inside declared protection zones near Perth. The instrument exists precisely for that: to restrict anchoring and trawling over the route. As a deterrent it works — that is our impression, not something the sources say — and both cables still failed inside them. A protection zone lowers the odds of something happening; it does not create an alternative route when it does.

The corporate equivalent is the service level clause. A penalty gives you money back; it does not give you the service back. A link with a flawless agreement can take days to come back if the fault is in the street works, and that penalty gets paid while the office sits idle.

No sensor, no cause — and the owner says so himself

One detail in Slattery's post deserves more attention than the ship chart: had those cables had FiberSense installed — a detection system SUBCO does use on other cables of its own — the cause of the fault would have been known instantly. Indigo West and Central do not have it, so the consortium will not know the root cause until the cable is recovered during the repair. Weeks after the break.

Translated to a corporate network: what telemetry changes is not the cut, it is how your response starts. Either you begin with a data point or you begin with a guess. We run our own network as a carrier, and this is something on-call duty teaches you: with NetFlow, per-link latency and loss probes and a public looking glass, you know within thirty seconds whether the problem is yours, your carrier's or a third party's. Without that, the first hour goes into arguing about whose fault it is. We wrote something similar about why a firewall will not save your access link: some problems are only visible — and only fixable — upstream.

Eighteen days

AusInternet, a regional provider that uses Indigo West, has been flagging the outage on its site since Saturday evening and is routing traffic around it. On the repair, the notice promises nothing it cannot deliver: the Indigo team has engaged its marine maintenance operator and the earliest expected fix is 26 August, with the caveat that "all timelines remain subject to clearance, logistics arrangement and weather conditions". From Saturday the 8th to the 26th is eighteen days — that subtraction is ours, and it assumes the expected date holds.

Some context, before anyone builds a theory. According to the International Cable Protection Committee itself, the world has around 500 submarine cable systems, total length has grown from one million kilometres in 2014 to roughly 1.7 million in 2025 (on the same page the ICPC also cites 1.8 million; we are quoting both), and 150 to 200 faults occur every year, a stable number despite the network's growth. Between 70% and 80% are caused by accidental human activity: fishing and anchors, not sabotage.

So 150 or 200 cuts a year is the planet's normal rate. The internet does not go down because of them, because there is another route — and the ICPC itself recommends that governments encourage "multiple and diverse" cable landings. It is the second word that almost nobody buys.

Now, your office

The scale changes and not much else. What repeats is the mechanism: two things that looked independent and were not. These are the things that turn two lines into one, ordered by how often we find them when auditing a site:

• The same wholesale network. Two different carriers, two different invoices, the same underlying access. It is the most common case and the easiest to check: just ask in writing.
• The same building entry. Two fibres coming into the building through the same duct, chamber or riser. A digger does not tell carriers apart.
• The same exchange. Even if the last stretch differs, both accesses may aggregate at the same node. A backbone cut there takes out both.
• The same router. Two lines terminating on a single box, with a single power supply and a single config someone edited on Thursday.
• The same power circuit. Flawless fibre redundancy, one UPS, and that UPS still running 2019 batteries.
• The same software. Two identical boxes on the same firmware fail for the same reason on the same day. That duplicates the hardware and leaves the risk untouched.

The questions worth asking

You do not need a study. These seven answers, in writing, tell you whether you have two routes or two invoices:

1. Which wholesale network sits under each access? The name of the network owner, not the retail brand.
2. Where does each line physically enter the building? Two entries, or one.
3. Do they share ducting or a chamber in the first hundred metres of street?
4. Are the media different? Fibre plus radio, or fibre plus 5G, cannot share a trench by definition.
5. Do they terminate on different devices, powered from different circuits?
6. What is the committed repair time and who performs it? The time, not the penalty.
7. When was failover last tested? With a date. A failover that has not been tested this year is not a failover, it is an intention.

What an SD-WAN solves and what it does not

An SD-WAN does three things very well. One: it continuously measures each transport (latency, jitter, loss). Two: it decides per application which traffic goes where and fails over without dropping sessions. Three: it applies the same policy across every site without touching boxes one by one. For a company with several branches and voice over IP, that is the difference between a dropped call and a call that degrades for a second.

What it does not do, and we should say it because we sell it: it does not create a physical route that is not there. If both accesses run through the same duct, an SD-WAN will give you a beautiful dashboard with two links turning red at once. And it adds a new component to look after, because an SD-WAN's management plane is a target in its own right these days: we covered that with figures from CISA's Known Exploited Vulnerabilities catalogue in an SD-WAN does not go down, it gets reconfigured.

When you do not need it

We are a carrier and we sell connectivity, so read this with an eyebrow raised: not every company needs route diversity. If your business can go four hours without internet without losing money, breaching a contract or a single customer noticing, a second fibre with its monthly fee is money badly spent. A 5G backup on the router, properly configured and tested twice a year, covers that scenario at a fraction of the cost.

The honest conversation starts with the number: what an hour of downtime costs and how long you can take. We broke that calculation down in RTO and RPO with no fluff, and it remains the only sensible way to decide whether the second route pays for itself. When voice is your sales channel, the ERP lives in the cloud or you have signed service obligations, the answer tends to write itself.

The short version

Pull up the contracts for your two lines and look for the word "redundant". Then go down to the comms cabinet and see where the cables come in. If they come in through the same hole, you know what you have, and no dashboard fixes it: two invoices and one route. SUBCO will have to lift a cable off the seabed to find out; you need a torch.

Sources (consulted on 15 August 2026): the Indigo West and Indigo Central faults, the definition of a shunt fault, statements from Bevan Slattery and the Australian Federal Police, the absence of FiberSense on those two cables and the AusInternet notice with the earliest expected fix on 26 August — iTnews, 11 August 2026. Number of cable systems, total length, annual faults, share of accidental causes and the recommendation for "multiple and diverse" landings — International Cable Protection Committee FAQ (updated 20 May 2026). The eighteen-day count is ours, computed from the published earliest-fix date, and that date is subject to clearance, logistics and weather. Neither SUBCO nor the Australian police have established a cause or attributed the faults to any vessel or deliberate act; neither does this post.