I trust never an internal watchdog of a micro.
I, however, do not like external puppies because the likelyhood of a runaway program toggling a port pin is far greater than a runaway making the "double write" required by an internal dog.

Since microcontroller developers are only experts in running software and supervisor developers are the only experts on power on behaviour and noise influence. And until today I have never seen, that both experts join together.
I agree, I would not rely on an internal power on reset.

the same number from a different manufacturer can cause a different working !!!
Especially if decoupling and/or other things are not totally up to snuff. I too, find that the MAX is the most noise immune (compared to the microchip equivalent; however, when we decoupled with 100nFdirectly across the chip both worked beautifully.

Erik