A.

2. 1 micro-second is 10^-6 seconds (not 10^-9).

G.

a) The proof that I gave in the original solutions just proves that
there exists at least one configuration of chips where #bad > n/2
where it is impossible to tell if there is a good chip.  Here is my
original (and correct) answer. (Sorry about the seperate file for the
Bug Report, but the original solutions file is no longer with us).

Assume there are g < n/2 good chips then any subset of g bad chips can
conspire by claiming that every chip in this subset of size g is
good. We do not care what the rest of the bad chips do. If we assume
that we can do every possible pairwise test (this is the best we can
possibly do) then we will end up with two subsets of size g with no
way to tell them apart.