Well, the first one is fairly evident.  You have a negative
capacitance in the device model.

The second is because small resistors are not a good idea
for spectre, because it leads to convergence difficulty
(put shortly, this is because of trying to ensure Kirchoff's
laws are maintained, and so when the normal numerical
methods are used to compute the voltages at the resistor
terminals, these would have to be done to excessive
precision in order to get the current across this small resistor
to be computed correctly to meeting Kirchoff's law). I
think this is covered rather more eloquently/clearly in Ken's
book, if I remember rightly.

Regards,

Andrew.