The parts you set aside were still in the bin.

On August 12, Rivian filed recall 26V508. It covers 3,829 of its 2025 and 2026 R1S SUVs and R1T trucks, over a front turn signal that can fail to light up. That is a real safety problem. It violates FMVSS No. 108, the federal standard for lamps and reflectors, and the reason is plain enough to picture: a vehicle that means to turn and has no way to say so, in front of a driver who has no way to know.

The detail worth keeping is that this is the second time. Last year Rivian filed recall 25V387 over the same failing turn signal. This one expands it. From June 2025 through July 2026, the company logged 17 more instances of the signal going dark. Then it worked out why more cars carried the bad part than the first recall had accounted for. The supplier had not properly quarantined the suspect parts. The parts everyone believed were pulled and set aside got installed anyway, into 3,204 R1S units built between June 12, 2024 and May 26, 2026, and 625 R1T units built between October 1, 2024 and March 19, 2026.

So the first recall named a batch. The batch was supposed to be isolated at the supplier, held back, kept out of the line. It wasn’t. The same defect shipped a second time, in nearly four thousand more vehicles, because the step that was supposed to contain it did nothing except get assumed.

Every shop runs a quarantine, whether or not anyone calls it that. The cracked tile you lean against the back wall to return, so it doesn’t end up on a job. The invoice you flag to fix before it goes out. The batch of trim you set on the far end of the bench because it came in warped. The customer file with a note that says don’t schedule until the deposit clears. Each of those is an isolation step. Each one depends on the bad thing actually staying separated from the good thing.

Here is where it fails, in a twelve-person shop the same way it failed at a car company. The isolation is a label, not a wall. The warped trim is a foot from the good trim, in the same rack, and on a busy Thursday somebody grabs the nearest piece. The tile to return never leaves the truck, and three weeks later it goes up on a wall because it was there and the crew was short. The flagged invoice sits in the same pile as the clean ones, and the flag is a mental note that belonged to the one person who was out that day.

A batch you have set aside is not contained until the bad units have physically left the place the good units come from, and someone has confirmed the count. “I pulled those” is a memory. “There were nine, I pulled nine, the rack is empty” is containment. The gap between them is exactly the gap Rivian is now recalling four thousand cars to close.

The cost of getting this wrong compounds in a specific way. The first failure is an accident. The second failure, from the same known cause, is the one that gets you looked at. Rivian caught the defect, named it, recalled it, and still shipped it again, and the second recall reads worse than the first because the problem was already understood. Your version is the callback on a repair you already did, the second warped piece from a batch you already knew was bad. Nobody forgives the repeat the way they forgive the original.

None of this asks for a system. It asks for one habit: when you catch a bad batch, finish the isolation instead of assuming it. Move the bad units out of reach of the good ones, and count what’s left.

When you catch a bad batch, is your isolation step something you have verified holds, or something you assume held because no one went back and counted the units still in the bin?

← More observations