We need to remember that ties rarely occur except in the case of very small elections, but it is just those very small ones where voters can see what is happening, and where the effect of later preferences upsetting earlier ones may be most troublesome.
In the real case quoted by Kitchener, there were 4 candidates for 1 seat. The 4 candidates were also the voters but not everyone voted for themselves. The votes were
ABC 1 BAD 1 ACDB 1 BCA 1giving an AB tie for first place whether judged by Alternative Vote or by Condorcet. Using Borda scores as tie-breaker, A is elected, but this is solely because of a third preference for A against a fourth preference for B.
Now Voter 2 has a right to be cross about that. He put A as second choice meaning, according to all the best explanations of STV, 'If B is out of the running, then I wish to support A' but B was not out of the running at that point.
Suppose there were the same set-up the following year. Voter 2 is likely to decide to plump because putting in a second preference the previous year was to his disadvantage. But Voter 1 may realise this and decide that he must plump too to counteract Voter 2's plumping - then Voters 3 and 4 will need to think about their strategies.
Whether anyone decides to plump or not is not really the issue. What matters is that tactical considerations have been allowed in, where STV (in its AV version in this case) is supposed to be free of them.
It may seem a pity to decide it at random, but such looking at the votes only decides it on the grounds that Voter 3 preferred D to B whereas Voter 4 preferred A to D. Is that really relevant when D is clearly out of it anyway?
My own conclusion is that to look at later votes in such circumstances, by Borda scores or any other method, is not a good thing to do, but I recognise that it is a matter of judgement, not of a clear right and wrong.