Coincident detection across separated stations: excluding chance and cross-talk

This report governs public treatment of detections reported at two or more separated stations inside the same review window. The method is intentionally conservative: simultaneity is considered a reason to slow the review, not a reason to promote a candidate.

The public copy preserves the chain of reasoning while removing station pairings, raw timing offsets, and security-sensitive facility data. It is suitable for citation when describing the Laboratories' false-alarm posture.

Each candidate begins with a time-base audit. Receiver clocks are compared against the station discipline record, network drift logs, and any manual service notes filed during the window. A candidate is rejected if the apparent simultaneity can be produced by clock correction, dropped packets, queue replay, or shared network behavior.

Infrastructure checks are performed before source discussion. Shared power events, cable work, temporary grounding changes, enclosure alarms, and remote-access maintenance are treated as ordinary causes until excluded.

After infrastructure review, the candidate is tested against retained station hours rather than against a simplified calendar. This keeps quiet periods, maintenance downtime, and inactive receivers from reducing the false-alarm estimate artificially.

The public report does not publish the full model matrix. It does publish the rule: a clean story is not accepted when a large enough archive could have produced one by chance.