I'm not yet convinced that the "reset chips" are any better than the naked MCU in cases where the risetime of Vcc is long.

I think, based mainly on my experience with prototypes, that the entire issue can be remedied by proper power supply distribution, i.e. a solid poured layer of Vcc and a solid poured layer of GND over the entire circuit board, coupled with a low-noise (very low ripple, very low load sensitivity) power source. Bypass capacitors must be of a low-ESR type, else it's not worth using them.

I like lots of capacitance on my supply rails, but this contributes to slow Vcc rise time on power-up. So far, I've had no reasonable success with SMPS or "weak" linear power supplies. When I use a linear supply with 100x the requisite capacity, I get a nice, solid behavior every time. When I use a commercial SMPS of 100x the requisite capacity, I have marginal reliability.

With any kind of power supply at all, the DS89C4x0's seem to work fine, as they have a built-in reset circuit. They even work with a little simple-switcher circuit with only a 20% capacity margin.

I don't spend much time on this interesting problem, since I use the DS89C4x0's. Eventually, I may be motivated.

RE