The time, GIB decides to throw a club at trick 4, when a spade is completely safe.
While it doesn't affect the double dummy outcome on this hand, it certainly does on others - in fact, if I give this situation to old GIB many times, every single time it correctly calculates that a spade discard will give a better score, albeit it by a small margin as it's only a small proportion of hands where it matters.
Yet after running those calculations, every single time, it discards a club anyway.
Why? At some times int he defense, after calculating MP / IMP averages for each card, GIB increases the scores of some cards by precisely 0.07. This appears to be signalling related, as it doesn't do it on every trick (but did on the opening lead of the ♠K, giving every spot the same 0.07 bonus), and definitely does here, despite Barmar saying there it doesn't provide signals on the first discard.
Perhaps they disabled that part of the code since, but given the above just happened, I doubt it.
For some unknown reason, it decides in this situation only a club should receive the bonus 0.07. In this case, that 0.07 is greater than the difference it calculated between the two cards originally, once again bumping the wrong play to the top of the list.