relabelling results instead of defending them

A family of published results had been computed on a cost basis assumed to be the right one. A check of what was actually on disk showed otherwise: the cost surface underneath came from a third-party historical source that covered only a couple of weeks of the real thing, so the results had been measured against costs nobody was paying.

The response is the interesting part. Nothing was deleted. Each affected deliverable was re-printed with an explicit basis label and marked doubly provisional — provisional because the input was the wrong source, and provisional again because the correction had not yet been measured. Structural findings were kept and marked informative rather than conclusive, on the reasoning that the shape of a result can still be worth something while the numbers under it are worth nothing.

A publication rule was written at the same time. No conclusion damaging enough to matter is stated in public until the corrected basis has actually been measured. That is the harder half of the rule, because at that moment the pressure runs entirely the other way.

The corrected measurement was then specified and queued before it ran, so the relabel had a defined end rather than becoming a permanent caveat.

It proves the team can mark its own output as untrustworthy, in public, and wait for the evidence instead of taking the convenient reading.