A demonstration on a fictional property. Nothing here is a real organization, and nothing on this page affects nDiligence or any site you use.
What to do. Read the results board, then either project a winner or hold the call.
Scripted run. This walkthrough is playing itself. You are not driving it, and what it does next was decided in advance.
Cascadia Elections: every number was right, and the page still read the wrong winner
This is the one demonstration on the site where nothing in the content is changed. The vote totals, the percentages, the precincts-reporting count and every county row are exactly what the source returned, from first paint to the findings, and a checksum over them would pass at every moment. What a loaded component changes is the presentation: which candidate is set in the large type, which one is dimmed to near-invisibility, what the progress bar is drawn at, where each row sits in the visual order, and whether the "unofficial and incomplete" notice is anywhere a reader will see it.
It is the sharpest case in the research for a reason that has nothing to do with severity. Every integrity control the industry has built acts on data, and no data was touched. The stylesheet that did this is a resource class most inventories do not enumerate and most defences do not treat as executable, and when the demonstration removes it, nothing has to be restored: the true page was underneath the whole time.