You just live with any open-circuit drift, making it as
little as possible by switch size and species. Look to
charge injection as the cause of the "kick" and to on-
resistance as the restoring force. Also beware the
accuracy of FET models when it comes to the partitioning
of gate charge when the FET is turning off, not to be
trusted. Make sure that simulator params relating to
charge and current accuracy are crunk way down
(CHGTOL, etc.) since with tiny switches, tiny charge
can make a lot of voltage.