I've been following this thread with interest.  One of my earliest PIC
designs (16c71) exhibited strange behaviour with slow power supplies - much
as has been described here.  The chip locked up with all outputs in a HIGH
state, which led to a potentially unsafe condition.  Since then, all but one
of my projects using PICs have had external watchdogs plus a timer which
monitors the watchdog and disables critical outputs if the watchdog trips.
I used to use the MAX705 but have been using a Xicor part for the past
couple of years.

The Xicor part (X25043) has a couple of nice things about it.  It includes:
power-up-reset, power supply supervisor, watchdog, and 512 bytes eeprom.
Its not too expensive (~3.00 Cdn) and has been utterly reliable.  One of the
nice features NOT in the data sheet is that the low voltage dropout detector
threshold value can be changed using a simple programming jig (it is not
possible to change in circuit).  This means that I only have to stock one
part number for the different supply voltages.

I mention that I disable critical outputs if the watchdog trips.  This is
just a simple 2 second timing network and comparitor tied to /MCLR via a
diode.  The timing cap is discharged anytime /MCLR goes low.  The time
constant is long enough to insure that the outputs remain disabled during
sucessive watchdog resets.

The X25043 that I use is in an 8 pin DIP package.  I use RB5,6,7 to talk to
the chip.  The rest of the circuit is arranged so that high voltage on /MCLR
will not cause damage.  The whole idea is that I can leave the watchdog off
the board until test time and use the watchdog socket as my In Circuit
Programming connector.  It has worked well for the last two and a half years.

Dwayne Reid   <dwayner@planet.eon.net>
Trinity Electronics Systems Ltd    Edmonton, Alberta, CANADA
(403) 489-3199 voice     (403) 487-6397 fax