Task: Promote and pull in sprint 23 stories from the backlog

• DQ/Refdata service boundary cleanup
• IAM/Refdata service boundary cleanup
• Sprint 22 leftover cleanup (absorbed 4 already-written task docs parked at sprint 22 close, unchanged)
• Refdata entity NATS event registrar audit (scope investigated live: 8/34 entities wired, not the 3 the capture originally named)
• Stochastic tick arrival times for synthetic feeds (merged with duplicate Vary synthetic feed tick frequency stochastically, which moved to discarded/; refined with a shared arrival-profile-config design and the GMM-vs-MMPP correction from review)
• Add a default party per account with a quick-login checkbox
• Audit refdata entities for composite (temporal-versioned) child relationships (net-new story scoping the generalisation the user actually wanted, after book_hierarchy_support.org turned out to contradict a later decision and was discarded instead)

• Book data model cleanup — misfiled in product_backlog/inbox/ despite sprint 22's table correctly listing it as STARTED.
• Resolve codegen model unification blockers and Codegen developer experience improvements — same misfiling; the latter's --profile=/–component= wording updated to --address to match the CLI reality landed by the former.
• Market data cleanup: retire dead duplicate tables — misfiled, never started; verified all 3 tasks still live in the codebase.
• Consolidate history dialogs onto HistoryDialogBase — carried from sprint 19; its Phase C gate (codegen org-model migration) completed 2026-06-11, well after the story was last touched, so it was unblocked and updated to say so.
• IR Rates synthetic data generation and Synthetic data collections: Basic and Realistic — both mislabeled ABANDONED in sprint 22's table with no corresponding decision recorded in the story doc; verified their remaining tasks were still genuinely open and carried them forward instead.
• Three ABANDONED-but-fully-reconciled stories moved back to their correct sprint_22/ folder (not sprint 23 — nothing left to do): Commission: counterparty, Commission: party, Commission: party_status, and Decommission ores.codegen.table model type — each verified task-by-task to trace cleanly to a sibling story or superseding capture; zero orphaned work.

• Model calendars as proper ORE Studio reference data
• Cross-rates matrix (CRM) (thin original acceptance criteria expanded into a concrete task plan following the domain model's concept map)

• Moved a stray capture out of doc/agile/inbox/ (wrong location entirely) into the real product_backlog/inbox/.
• Corrected an earlier mistake in this same session: a discarded duplicate capture (Vary synthetic feed tick frequency stochastically) had been git rm'd outright instead of moved to discarded/ with a reason note — fixed in a follow-up commit.
• Discarded Sweep pre-template-fix entities: add security definer and active-rows bootstrap filter as fully resolved by natural recommissioning (verified against generated SQL for all 10 entities it named).
• Discarded Add net-new self-referencing hierarchy support to book as superseded by a later, more considered decision that book-to-book hierarchy is deliberately not a feature.
• Stripped shell/CLI implementation scope from 5 near-identical commission_* template stories still in inbox — that work is now owned by the top-level Shell/CLI commissioning stories, not per-entity ones: Commission: business_unit, Commission: portfolio, Commission: business_centre, Commission: party_id_scheme, Commission: contact_type.

Extended agile-refine-backlog with a deep refinement mode (see Deep backlog refinement, the sibling task that will exercise it) — added to the existing skill rather than creating a new one, per explicit instruction to avoid skill proliferation.

Rebuilt the compass search index (compass index --rebuild --org-roam-db-sync) once mid-session, and the documentation site twice (the first rebuild raced with in-flight file moves and failed; the second, after all moves had settled, succeeded).