Our simulation software has a fault coverage mode in which it ties every node both low and high, and determines if the simulation results change. It the results don't change, then that fault was undetected, and the software reports the node.

We have a saying that "if you haven't looked at it, then it's broke."

Remember, I design 8051 chips, so I have a netlist with every internal gate.

--------

We also check for excessive Idd. There are a number of internal faults that only show up as excessive Idd. In the case of your SRAM, leakage on a bit line can pass a quick production test, but fail in your system as the part warms up with use.

It is our policy to scrap "good" parts whose only failure is Idd. It is far cheaper to do that than to risk a future return from a unhappy customer. Consider how long some people have remained mad at Atmel.