Home internet modem keeps powering down and restarting

Um. I'm an electrical engineer. I troubleshoot hardware for a living, at least until recently when I retired.

Brand new T-Mobile Gateway, the tall, square in cross-section one. Started seeing $RANDOM dropouts, once every couple of hours, as soon as the thing was put in use.

On day, happened to be in the office upstairs where the thing was placed, down went the internet. Looked over, screen says, "Powering Up". There had been no power outages whatsoever. When it came back up, so did the internet.

Four hours later, lather, rinse, repeat. Called T-Mobile support. After the usual five minute wait, talked to a nice service rep.

Se said it was clearly a problem with the gateway. AND that that particular brand (the tall one) was known to have this problem. Drop shipped me a new one on a Friday, got it on this last Monday. Opened the box: Same rough shape (square in cross-section) but shorter than the original. Set it up. No more drop-outs.

Did have to swap the SIM card from the intermittent one to the new one, but that was it, beyond the usual setup follies.

Could be the power brick. But, speaking as a troubleshooter: I've seen hardware that does stuff like this. A fair number of reasons. Overtemperature on some device that causes excessive current draw; reset sensor whose voltage thresholds are too high/too low, so it causes a reset; bad parts in manufacturing that short out intermittently; and so on. I note that it took a while before it started doing this in my case, at least four or five hours. Most factories power up their equipment, sometimes in a heat tent, with the intent of detecting early failures. So this looks like a factory escapee: Should have been caught during testing, but worked long enough to ship.

There’s this general idea: One can do a mucho of testing when one builds a product. Testing the circuit board, lot testing individual components, testing the assembled board before putting it in the box, testing the complete box after assembly (functional test), and, of course, heat tank testing. If one is getting 99%+ yields at a particular test step and there's decent fault coverage, it's often cost-effective to skip some tests and use others. This works if one's suppliers provide low-fault batches of hardware; one pays more for that, but then gets to skip hiring people and mounting test stations, so it's an interesting balance for the manufacturing engineer.

But there are pointy-haired bosses who just love skipping testing steps in the interests of low costs.. and this smells of that.