That would potentially delay system startup by a few milliseconds. Is that a problem?
NO

2. I can put a voltage detector that will force a hard reset at Vdd = 4.25 (or whatever). On the downside, this might force the processor into reset under noisy conditions (such as whenever the motor turns on) (or worse, whenever the flash is programmed!). To compensate, I could require the detector to have a low voltage signal for 3 (or N) clocks before activating reset. Still, would anyone here authorize a second source for an existing design with that level of reset change? Do you trust your older systems to have a clean supply?
here it's fuzzy because "second source for an existing design" makes it tough. However in a case of second sourcing for somebodys (nonimpressive) existing design I found it to be required to solder a cap across Vcc and Gnd for a chip, so, if you run into that one you will not be the first.

3. I can make a configuration bit set the polarity of the reset pin. I suppose that the bit should also change the pulldown resistor into a pullup resistor. That would mainly be for new designs (I can always hope). So would you use an active low reset if it was available?
kind of a moot point. If your internal reset WORKS, who cares about reset polarity.

mainly be for new designs (I can always hope
if you ever get close to RTP, come see me.

Erik