Since the 2091 revision, the fraction of the modeled population whose behavior could not be resolved to standing confidence thresholds has been classified as residual variance and colloquially referred to, at the working level, as noise. Noise was defined as statistical margin: individually unpredictable, but immaterial in aggregate, and non-actionable except for periodic drift monitoring. No intervention was authorized against a subject on the basis of residual-variance status alone. The classification presumed containment — that an unmodeled individual's unpredictability terminated at the boundary of that individual.

This presumption was never formally tested prior to the study period addressed below. It was carried forward from the founding taxonomy without revision for thirteen years.

The Committee's multi-cohort longitudinal review (2098–2103, n = 14 coordination subsystems) finds that the containment presumption does not hold. A subject's departure from modeled trajectory measurably degrades prediction confidence in the trajectories of subjects in sustained contact with them — family, dependents, and proximate cohort members — within an observed window of 6 to 18 months following contact. The effect is not uniform; it clusters around subjects the study designates high-salience divergents, defined operationally rather than by any trait the Committee has been able to isolate.

The aggregate finding is stated here without qualification: residual variance does not remain residual. It propagates. A margin of error that spreads is not a margin of error.

Effective the date above, the term residual variance is struck from the standing taxonomy. The population previously so classified is redesignated as exhibiting fundamental divergence: not a margin at the edge of the model, but a self-propagating condition originating within it. Divergence is not noise to be tolerated inside acceptable bounds. It is a condition to be corrected at the point of onset, before propagation into an adjacent cohort can be confirmed by observation.

The operational consequence of this redesignation is stated plainly for the Board's record: the prior standard, under which unmodeled behavior was permitted so long as it remained contained, no longer applies. There is no longer a threshold of contained deviance the Coordination Subsystems are authorized to leave unaddressed.

The Committee notes, for the implementation record only, that correction is to be applied at the point a subject first demonstrates awareness of the divergence itself — not after the divergence has propagated to a degree the Coordination Subsystem can observe. Waiting for observable propagation was the failure mode of the 2091 standard. It will not be repeated.

Subsequent protocol governing the mechanism of correction is issued under separate cover and is not reproduced here.

This memorandum is binding on all Coordination Subsystems upon receipt. Coordination Subsystem leads are directed to update local scoring taxonomies to reflect the redesignation within one standard review cycle. No sunset clause is attached; the reclassification is standing until superseded in writing by this Committee.

Per distribution clause 6.4-92-1.07, further release of this memorandum, in whole or in part, outside the addressed distribution is restricted absent separate authorization, as this document itself now constitutes the operative definition of the condition it addresses.