Low cost MCUs have FLASH which has a finite life. Same with FPGAs which need to have their brains filled from something.

I worked on a project a few years ago that had a PIC part. They never got the software working right so we looked at the original spec. The entire software could be crunched down to 7 gates and a comparator. So we did a rev2 board with a flip flop and a NAND and a LM358N and there hasn't been a failure since. To be honest with some bastardising to get hysteresis out if a 555 and a couple of transistors would have done as well.

The BOM was £0.09 more expensive but the part never failed in production.

The most interesting thing is that PIC part no longer exists so there is no service option other than replace which costs more down the line.

(This was a watchdog monitor for signalling systems)

Edit: yes I know the 358 isn't a comparator but it was fed just before saturation to get hysteresis avoiding more parts.

Those are important considerations , but:

You did the design a few years ago, before the silego was a good option.

But for many design , esp in consumer , the silego part would be a good fit because it would work as same as discrete parts,it's manufacturing life and reliability(12 years) would be good enough, and it would be better in other parameters.

And once the consumer market is taken, we might see versions targeted at more demanding applications. But it's not certain.

12 years isn't that long. That's less time than XP was around to put it in perspective...

in consumer markets, do people design for lifetime over 12 years ? Because my feeling about consumer market that it's mostly crap, reliability wise.

Rarely. They want it to get through the European statutory warranty and that is it.

I have an oscilloscope built in 1976 by HP and it's showing no sign of any problems yet. I'd like to see more stuff like that on the shelves.