I insist on what I've written above - you picked up an infinitesimally small subset of possible causes of failure and want to invest an inappropriate time to test it.

From the point of view of $2500 per board and the chance of its loss, I'd concentrate on the passives and discretes and linears - they are easier and cheaper to test anyway. I'd try to identify the most smoke prone parts of the circuit (e.g. supply) and measure at least the nominal vale of 100% of parts going there.

Also, on the '51 iself, you intend to perform only a logic test (notabene using the device under test as a test controller). IMHO the most likely mode of failure once they left the manfacturer's site is I/O corruption due to improper handling. Your test would cover this only incidentally.

So I think you should start with developing an analog test rig to be able to set and measure voltages and currents on the I/O pins. Some simple software is needed, too, to wiggle with the output levels.

Also, ICCQ is not a specification, but a test, where the quiescent (static) supply current is measured differentially between all possible combinations of states. This is aimed to discover typical "hard" faults in CMOS structure. This test potentially discovers more faults than a "digital" test, as the latter might pass marginally even if a fault is present.

Anyway, one of the key problems is to generate an exhaustive enough test vector. This is a hard task even if you know the internal structure, but next to impossible if you don't. There are practically infinite combinations of all states even in a simple 8 bitter and you cannot test them all in nearly reasonable time; so you have to make assumption and simplifications, which arise from the structure (layout and function).

Testing is paranoia - and you cannot be paranoid enough...

JW