AgeMOS model for reliability simulation
If you have to use a process that has significant aging
effects (I refuse), -and- you can get both quality models /
data for fitting, and certainty that the aging behavior is any
kind of consistent, then modeling what you have to use is
about the only thing you can do.
But in my experience, such as it is, consistency (lack of)
is the real reliability problem.
I'm sure periodic retuning will work, at least until drifts are
beyond the trim range. Seems like merely an extension
of autozero methods, more sparse. But you have to question
whether your drifts will really be confined to the analog
front end, or so pervasive that trimming the front end alone
will not make the larger circuit that much longer-lived.