Yes, this is a commercial product (or will be when my customer installs
them in his module). There is quite a substantial ammount of other testing
complete on these, thats how I caught the power glitch event (which occures
extremely rairly right now).

  No, there is hardly any room left, I might squeeze in a SOT-89, thats all
(I'm using 20x40 mil resistor chips, those fit easily but the traces to
then are the problem).

  The Dallas Semi power monitor is nice, cept it has an internal time-out
on power up. The application is a timer that times from power up, hence
anything over a few tens of microsecs is prohibative (and will interfear
with my timing accuracy, something that is importaint here).

  I've revised my first few units to have a bleed resistor (actually
already included on the board for a slighly different model timer) and they
pass the tests I'm specfied for (ignore a .5ms interupt, reset for a 50ms
one). A power glitch at some specific time in between those will upset the
PIC, but this should be a non-event as the power line driving it is itself
specfied not to have such a glitch.

  What I think will be my definative test is to find the bouciest switch I
can (like a Radio Shaft power pushbutton) and drive it off that. If it
survives the bounce I should be OK and will be able to return to restful
sleep.

  Thanks all for the comments.