Throw away the cable that came with the phone

Early on, customers wired up their devices with the cable that came in the phone's box. Reasonable — it's right there, it's the manufacturer's own, and it works. Then the lab starts doing the thing that costs the most time in this business: failing intermittently, for no reason anyone can see.

A device drops off its node mid-test. A build transfer that takes twenty seconds suddenly takes four minutes. A phone that is plugged in, charging, and visibly on simply isn't there as far as the node is concerned — until you unplug and replug it, and then it is. Every one of these looks like a flaky test, or a flaky device, or a flaky platform. Almost none of them is. The cause is thirty centimetres of copper nobody thought to suspect.

Why the cable is the worst kind of failure

It's invisible. When a test flakes, you look at the test. When a device misbehaves, you look at the device, the OS, the automation. The cable is the last thing on the list, because it "obviously" works — it charged the phone, didn't it? So you spend a day ruling out everything expensive before you get to the cheapest part in the rack. It's the same trap as "it returned OK": the signal you're reading says success while the thing you care about quietly failed.

Cables are not interchangeable, and the industry knows it

The idea that a cable is a cable is simply wrong, and there's a famous demonstration of how wrong. In 2015-2016 a Google engineer, Benson Leung, ran a public campaign reviewing USB-C cables for spec compliance and found a startling number that weren't — including one miswired cable that physically destroyed the laptop he was testing with, plus two of his analyzers, because its ground and power pins were crossed. That episode is a big reason the USB-IF runs a cable-certification program at all; its own page exists because, in plain terms, "all cables do not have the same capabilities."

That story is the dramatic extreme — a dangerously miswired cable is rare. The failure a lab actually lives with is the quieter cousin: a perfectly ordinary cable that is marginally out of spec, or was fine on day one and degrades. Which brings us to why in-box cables in particular don't survive a lab.

A consumer cable is not built for a lab's life

The cable in the box is engineered for a person who plugs in once a night. A lab does something completely different to it: connect and disconnect thousands of times, hold a stable data link and charge current for hours, run 24/7, for months. Connectors wear. Marginal conductors that passed at room temperature on day one start dropping the data line under sustained use. And because in-box cables aren't marked with which spec they meet, you have a rack full of subtly different unknowns — the exact fragmentation problem, shrunk to the smallest part in the system.

So we don't use them. We ship known-good, tested cables as part of the lab and treat them as a controlled, consumable part — the same make, to a known spec, replaced on a schedule rather than when they fail. It sounds almost too small to matter. It removes an entire category of "the lab is flaky" tickets, because the question "is it the cable?" is answered before anyone has to ask it. The cheapest component in a device lab causes a wildly disproportionate share of its failures — precisely because it's the last one you suspect. Control it, and a whole class of ghosts goes away.