Court of appeal

Suppose the officer's diagram suggests Jonah entered late. The traffic video is blocked. Priya's statement and the second witness support the report. The case appears settled.

Then Tomas's phone provides a photograph taken seconds after impact. It shows the pedestrian signal in a phase inconsistent with the report's timeline. The bicycle data places Jonah in the intersection earlier than assumed. The store video, retrieved before deletion, shows the van beginning its turn while the bicycle is already visible. No single trace becomes the master file. The strength comes from convergence across registers with different routes of error.

Jonah's internal account meets Tomas's social witness. The bodily injury meets the physical mechanics. The institutional report meets digital timing. Priya's account is not discarded; it is reinterpreted in light of the blind spot and time pressure. This is distributed verification.

An archive becomes corrigible when one register can challenge another.

A healthy court of appeal allows memory to meet record, record to meet witness, witness to meet measurement, and official categories to meet evidence they did not create. It also allows later correction to enter the archive clearly enough that downstream systems do not keep retrieving the error as if nothing happened.

The accident will never be stored whole. Jonah will remember one version. Priya another. The body will keep its own weather. The report will compress. The files will persist according to policies no one at the intersection chose.

The task is not to build a perfect archive. It is to prevent any partial archive from becoming unquestionable merely because it survived.