02 — Count Reliability Label Model + Baseline (Phase A)

A1 · v_rp_count_reliability_label_model (controlled vocabulary)

15 labels with rank, semantics, ui_severity, default_confidence. A label dictionary is a controlled vocabulary, not a business count, so it is the only place literals are used.

label	severity	default_conf	meaning
FINAL	green	95	ratified/canonical/authoritative
CANDIDATE	amber	60	proposed/derived, not ratified
PARTIAL	amber	55	source partially covered
UNKNOWN	grey	20	no ratified def; surfaced NULL not faked
STALE	amber	40	snapshot lags live
SYNTHETIC	amber	50	axis/group in-view, not registered
ADVISORY	grey	45	aggregate/equation, not a pivot
NOT_ADDITIVE	red	40	never sum into a headline
OWNER_BLOCKED	amber	50	awaiting object owner
AUTHORITY_BLOCKED	red	50	awaiting president vote
NEEDS_LAW	amber	35	counting law undefined
NEEDS_ADAPTER	amber	30	source class uncovered
NEEDS_RECONCILE	amber	45	parallel sources disagree
DEPRECATED	grey	30	superseded/retired
TEST_ONLY	grey	25	test fixture, exclude from prod

A2 · v_rp_count_reliability_baseline (21 LIVE metrics)

Every main RP count is a live scalar subquery (no hardcoded numbers) plus: source_scope, source_object, source_freshness, count_semantics, reliability_label, confidence_score, warning_flags(jsonb), blocker_type, next_action.

Selected rows (live at run):

metric	value	label	blocker
rp_node_universe	87	FINAL	none
base_pivot_active	37	FINAL	none
base_pivot_all	39	NOT_ADDITIVE	lane_split_before_headline
base_pivot_test_only	1	TEST_ONLY	exclude_from_prod
base_pivot_deprecated	2	DEPRECATED	—
proc_official_rp	0	AUTHORITY_BLOCKED	president_vote
proc_candidate	19	CANDIDATE	—
proc_candidate_genuine	6	OWNER_BLOCKED	object_owner
pxt_nodes	12	SYNTHETIC	register_axis
trigger_live_db	408	FINAL	none
trigger_registry_table	107	STALE	rescan_or_reconcile
event_active / event_total	30 / 52	FINAL / NOT_ADDITIVE	—
dot_total	309	NOT_ADDITIVE	lane_split_before_headline
dot_executable_confirmed	186	FINAL	none
residual_reconcile_live	2	OWNER_BLOCKED	object_owner

Completion: every main RP count is labelable with reliability semantics. ✔