Some things like the co-variances of TOXP & TOXN,
deltalP and deltalN, u0P and u0N, VTP and VTN are
going to elude you if looking at WAT limits alone, and
using worst case for each may be unrealistically bad
(some are anticorrelated in part).
That said, you could -assert- some conditions and
let "whoever" criticize the details of that choice.
Like VT +/- 20%, Tox +/- 10%, Leff +/- 10%
(whatever deltaL makes that so) and so on.
Contemplate, add "kxxxx*" factors to the relevant
SPICE model cards, make something up for the "k"
values and loop it nested, then sort for the worst
you encountered.