[phkahler]

>>Is there an algorithm that would find the minimum subset of these to cover every parasite/bug?

This is the same problem we have in embedded systems diagnostics. We used to make a list of failure modes, and then a list of diagnostic tests. A given test may detect multiple different failures and I've made grids exactly like this. I always advocate for finding the minimum set of tests to detect all failures (to minimize false positives and simplify software). I was going to make the same point in this case, but parent already did.

To be clear, sometimes it's sufficient to detect failure without knowing exactly what the problem was. For specific diagnostics you may want POST to handle that, but at runtime you may only need to detect that something has failed.