... where something similar can be justified.

Generally, the manufacturers perform a relatively thorough testing, designed to reveal even those faults which don't result in faulty program execution under normal operational conditions, and they toss away those parts which fail these tests, so what you get is 100% operational (OK, Mr.Murphy might play his little wicked game, sometimes... ;-) ) under the nominal conditions. Also, the technological steps in manufacturing and packaging a chip are designed so that they ensure a certain expected lifetime (or failure rate, whatever you want). So, in most of the "normal" applications, you might be quite confident that even without a POST or such you are reasonably safe.

However, I can imagine a range of applications, where ocassional uncontrolled excess stress is not quite excluded, yet the device shall perform (or fail) consistently. For example, in an outdoor equipment it is not unreasonable to expect excess stress from a nearby lightning; and I can imagine a partial FLASH memory failure because of that.

On the other hand, we can also discuss, whether in these cases a power-on test is adequate enough. I would say, once you take the pain to do ANY self-testing, you can as well do it all the time, for example as a background event.

There are also applications and areas, where such testing - of RAM, ROM, peripherals etc., upon powerup or periodic - is required, to get certain approvals, certificates etc. Then of course there is no point of discussing whether this or that has a rational background or not...

JW