Conditional on O being completely non-viable, I think it’s strategically optimal to vote A>O>B>C (or A>O>C>B) whenever you’d bullet vote for A in an Approval Voting election, which includes both of the scenarios you described for ranking O second. If O can’t win, the question of deciding whether to vote A>B>O>C or A>O>B>C is simply a matter of whether B having more points is good or bad.

I agree with you that Borda Count doesn’t really require more coordination than Condorcet to result in the election of a universally hated candidate, where the amount of coordination is measured by the number of dishonest ballots. However, I think that getting to that level of coordination is far easier under Borda Count since in Borda, unlike Condorcet, strategic voting is effective even with an arbitrarily low number of strategic voters.