Think of ESD in terms of the current loop, not the pin
voltage. There will always be a return. Deducing what
path it will take, is all on you as a designer. You must
know the (often unadvertised / modeled) breakdown
attributes for all devices touching the loop, because
they may change that loop, and this includes "devices"
that most circuit designers will prefer to ignore (well
diodes, substrate-well-drain BJTs, closely spaced
conductors (X,Y,Z) as spark gaps, conductors as fuses.
If you had a desire to be thorough you would create
supplemental models that have "live" and accurate
breakdown response modeled (likely a subcircuit) and
do any-any pin zap simulations. But this entails a lot
of probably-not-appreciated char, modeling and design
work.
This is how I have approached pin protection design over
the last decade or so, anyway. If you are given I/Os to
use that have advertised capability and constraints and
requirements, then maybe you just follow the rules. But
custom analog never gives you much of that.