The Vessel Ownership Layer Has Been Frozen for 52 Days. The 843 Vessels With Obfuscated Identity Edges Don't Show Up Anywhere Else in the Graph.

Eight hundred forty-three vessels carry an obfuscation-flagged identity edge — an MMSI change or a shell-company hop — somewhere in the temporal graph. In the last 30 and 60 days respectively, none of them generated a course-alteration anomaly event and none of them appear in a single pairwise vessel encounter. Not a reduced rate. Zero.

The Setup

That absence matters because of what else is frozen. SHIP_MANAGER and OWNED_BY — the two edge types that record who actually operates and owns a vessel — have not added a single record since May 12. Fifty-two days, as of this writing. A third identity layer, flagged_as (flag-state registration), is not frozen: it added 578 new records in the last seven days alone. So the graph isn't uniformly stale — the specific layer that verifies ownership is stale, while the layer that just tracks which country a vessel is flagged under is actively updating.

The Chain

Put those two facts together and the shape of the problem changes. If SHIP_MANAGER/OWNED_BY were merely lagging, a vessel that changes its MMSI could still be cross-checked against its recent encounter behavior or kinematic pattern — did it start behaving differently around the same time it changed identity? For these 843 vessels, that check comes back empty. They don't appear in course_alteration (200,427 such events fired across the fleet in the last 30 days, averaging a 97.2-degree course change) and they don't appear in pairwise_encounter (2,470,895 recorded encounters across the fleet in the last 60 days). There is no independent behavioral trail to corroborate or contradict what the identity graph says about them.

Two explanations are consistent with this. One: these vessels are genuinely inactive — the identity change coincided with, or caused, an exit from tracked AIS behavior, which is itself the more concerning read for compliance purposes, since an MMSI change immediately followed by disappearance from every other tracked signal is close to the textbook obfuscation pattern. Two: an ID-mapping gap between temporal_edges and the newer event/encounter tables means these vessels are active but invisible to the join — a data-plumbing issue rather than a vessel-behavior one. The entity ID format is confirmed compatible across all three tables (a spot check found a shared vessel ID appearing in course_alteration, pairwise_encounter, and temporal_edges simultaneously), so this isn't a schema mismatch — but a compatible format doesn't rule out a coverage gap in which tables get populated for which vessels.

For scale, the ownership freeze sits inside a broader pattern of tightening encounter geometry that has nothing to do with identity obfuscation specifically: across 2.47 million encounters in the last 60 days, open sea produces the tightest average minimum range (0.471nm) of any context in the dataset — tighter than approach channels (0.530nm) and tighter than anchorages (0.586nm), despite open sea being the context with the most room to maneuver. The trigger mix behind those handoffs is stable regardless of context — extremis accounts for 74-76% of Rule 17 handoffs whether the encounter happens in open sea, an approach channel, or an anchorage. Context changes how close vessels get before the protocol engages; it does not change why it engages. That stability is what makes the total absence of the 843 obfuscated vessels from this dataset notable — the encounter-generation process is consistent and high-volume everywhere else in the fleet, and it still produced zero hits for this specific cohort.

The Implication

A frozen ownership layer is a known gap — you can date it, you can watch for it to resume. An identity-obfuscated cohort with no corroborating signal anywhere else in the platform is a different kind of gap: there's no fallback check. For sanctions screening or compliance workflows that lean on Overwatch's broader signal set when ownership records are stale, this specific 843-vessel population currently has no such fallback. The identity graph says something changed. Nothing else says anything at all.

What to Watch

Whether SHIP_MANAGER/OWNED_BY resume and, if they do, whether the new records backfill the 52-day gap or simply start counting forward from the resume date — a resume-without-backfill leaves this exact population permanently uncorroborated for that window. Separately, whether any of the 843 vessels generate a course_alteration or pairwise_encounter event going forward — a first hit would help distinguish "inactive" from "invisible to the join."

Limitations

The zero-appearance finding is a negative result, and negative results are the easiest kind to get wrong. It is possible some of these 843 vessels are active under a different ID than the one recorded on their temporal_edges record — identity obfuscation could itself explain a failed join, not just a genuine absence from tracked behavior. Confidence-decay values on shell_hop edges (avg 0.183, the lowest of any relation type) look like a trust signal, but the exact semantics of that field — whether low values mean "little decay has occurred" or "confidence has decayed to a low level" — aren't verifiable from the table alone, so no claim here rests on that number meaning one thing rather than the other. The course_alteration and pairwise_encounter baselines are healthy and high-volume, which is what makes the zero credible as a finding rather than a data-availability artifact, but it does not rule out an ID-mapping gap specific to obfuscated vessels.

Data as of 2026-07-03. Sources: Overwatch temporal_edges (24,770 edges), axiom_events (course_alteration, 30-day window), pairwise_encounter (60-day window, 2,470,895 encounters).