I have now performed experiments with parallel plates. It is clear
that there is a transfer of cherge from the metal plate to the
dielectric surface which takes place during the charging process.
One can see and hear the sparks responsible during the charging
process. The great majority of the free charge on the assembled
capacitor resides physically on the surface of the dielectric.
There is no evidence that any charge is transferred from the plate
to the dielectric during the separation process. No discharge is
seen during separation, and after separation the removed plate
still retains a considerable electrostatic charge (excepting the
case in which that plate is grounded).
I will suggest that the cookie tin experiment I described not be
attempted by anyone. Today I managed to generate a potentially
lethal charge using only a hand cranked Wimshurst machine. Other
tricks were necessary, but these are best left undisclosed. I have
discussed the ethics of this with three colleagues and they concur.