[Kalium]

You're absolutely right! Those calculations can be done without actual testing. The only wrinkle is that doing so comes with the risk of failing to foresee something or otherwise being wrong. So there are some error bars involved.

What kind of error rate are you willing to accept in models of pharmaceutical shelf stability? Bearing in mind that errors potentially translate into deaths, probably disproportionately of the less privileged among us?

[TheKarateKid]

I'm not sure how it would be any different than existing expiration dates for OTC medicines, food, dairy, etc. In the case of medication, it is most likely to become less effective over time rather than toxic.