a claim carries its status

In a record assembled by many hands, the dangerous line is always the summary. A claim that holds only under an assumption, once raised a level and restated, reads as though it holds outright. The operation's defence is boring and effective: the top line carries the weakest condition it still depends on, and it moves only when that dependency closes.

The history contains both directions of the event, and records them separately. A condition stands for many cycles, and the summary restates it every time rather than dropping it for neatness; then one commit retires it, names the work that closed it, and credits the lane that did the closing. Neither movement is a silent edit, and the first is what makes the second auditable.

The tension is asymmetric. A conditional headline understates what may already be true, while an unconditional one is a claim somebody can be held to. The summary is where overstatement is cheapest to write and most expensive to defend, which is why it is the line most likely to drift in a large operation.

The rule adopted is to understate until closure and then update once, with the reason attached. That means the index is periodically wrong in the conservative direction. Being wrong conservatively is legible and correctable; being wrong optimistically is indistinguishable from being right until somebody checks.

Cost: summaries lag the work, lanes must record dependencies they would rather not name, and every retirement needs an owner. What it proves is that the operation treats its own headline as a claim requiring a witness, which is the same standard it applies to everything underneath it.