Actually, the measured parameters (or better, parameters derivated from many measurements under various carefully controlled conditions(*)) are input to the circuit modelling software...

The IC making process is quite complicated and has many variables involved. There exist process simulators and the parameters of the transistors can be simulated from first-principles too, but at the end of the day, after the process is designed and installed, first testing structures are manufactured and are measured extensively, to determine the real model parameters; and even afterwards, the manufactured wafers contain testing structures which are measured again and again to ensure the consistency of the process.
Oh yes, and what I am talking about is what I learned in school 15 years ago; since then the technology has been scaled down by about a factor of 10, which certainly brought up additional effects into the first-principle simulations... I don't think the testing will ever be abandoned.
Oh yes, and the evaluation of reliability... That's nothing more than extensive testing again - see reliability reports which are published for many chips.

-----

(*) I think this is the point where the "testing" which Erik is agains turns into science - one has to consider many things when making the "test" - and the OP would probably not have the experience nor instrumentation to do it properly, I admit. Nevertheless, I think it IS possible, but not easy nor cheap to determine how a transistor can be zapped and if it will have any impact to the remainder of chip (including the reliability). It might turn out that it can be zapped as well as that it cannot be zapped. But saying: NO it's impossible... Well if one doesn't state the boundary conditions, one can obtain any answer one may wish.

As I said, if I would have those diamonds... How much is an army today?

Jan Waclawek