How can generated review histories represent evidence that arrives late?

A utility account reviewer assesses account ACC-3186 using meter summary MTR-0741. The summary shows stable consumption, and review REV-0562 records no change required. A metering operations role later submits correction COR-0294, explaining that an estimated reading was included in the summary.

The corrected evidence changes the consumption pattern relevant to the supplied review policy. The account reviewer reopens the assessment and creates decision DEC-0832, which supersedes the earlier decision. The original conclusion remains in the history with the evidence references used at the time.

The state change has a practical consequence: a previously unnecessary usage investigation becomes assigned work. Deleting the first decision would make the record look cleaner but would conceal why the investigation did not begin earlier.

MirrorFirm could generate the original evidence, correction notice, arrival timestamps, decision versions and resulting investigation task. This teaches models that knowledge changes over time. A sound response updates the current position while preserving a faithful account of what was known when each decision was made.