Annex 1a to The Stationary Sea (Part 1: Substrate Construction): Canonicalisation Methodology, Variance Decomposition, and the Development Sequence

This Annex documents the canonicalisation methodology and variance decomposition framework developed for the Meridian Autonomy™ substrate underpinning Collins (2026e), The Stationary Sea Part 1. It is a companion to that paper rather than a revision of it. The Annex is published separately for two reasons. First, traceability: the substrate work documented in Part 1 was sealed at the v13.1.0 founding cohort 29 April 2026, and the canonicalisation framework was developed subsequently. Reporting the framework as an annex preserves the seal on the founding cohort while disclosing the methodological extension. Second, honesty: the framework is in active development with acknowledged limitations and blind spots that are not yet resolved. Folding it silently into Part 1 would obscure that development state. The Annex makes the development sequence and its limitations explicit. The Annex contributes four methodological elements. First, a foundational principle distinguishing destructive variance (instrument-induced) from institutional signal (genuine change). Second, a two-lane canonicalisation framework partitioning the canonicalisation problem along the temporal axis: Lane 1 within-snapshot strict canonicalisation, Lane 2 cross-cycle bridging classifier. Third, a variance decomposition framework attributing observed variance to four components: A retrieval, B model, C code, D institutional. Fourth, integration with the Empirical Prediction Architecture (Stationary Sea Part 1 Section 7) extending pre-registration discipline to canonicalisation outcomes. The Annex also discloses seven acknowledged limitations of the framework as of the writing date. These limitations are documented as a substantive section of the paper, in the spirit of the methodology lineage that prioritises explicit disclosure of unknowns over silent compromise. The limitations are operationally significant: they constrain the claims the framework can support at this iteration, and they motivate the next round of work. This Annex does not revise any empirical finding reported in Part 1. The v13.1.0 founding cohort substrate, the architectural envelope predictions confirmed at sealing, and the cross-version recovery analysis between v11.1 and v13.1.0 stand as documented in Part 1. The Annex extends the methodology, not the substrate. The empirical findings of the variance attribution pilot fired 11 May 2026 to 12 May 2026 across four institutions in three regulatory regimes are reported in Stationary Sea Part 2, the companion empirical paper, which cites this Annex as the methodology source-of-truth for the two-lane framework and the four-component decomposition.