Subset of skippedClasses that are NOT IfcRoot descendants — records
with no GlobalId (geometry, placement, and style resources). This
bucket fires on essentially every real IFC file and is not, by itself,
evidence of a problem.
Always true when this object exists. Distinguishes "the census ran
and found nothing" (an object with ran: true, totalScanned > 0,
empty drop lists) from "the census never ran" (the field is absent) —
callers must check for the field's presence, not just its contents.
Count of distinct IFCREL* classes routed into a relationship-graph bucket.
Count of distinct IFCREL* classes seen on the wire.
Subset of byClass with category skip.
Subset of skippedClasses that ARE IfcRoot descendants — a class
with its own identity (GlobalId) that nonetheless never entered the
entity table. This is the subset worth alarming on: it is the shape of
the #4208 IfcCovering-fell-out-of-GEOMETRY_TYPES incident
(documented in columnar-entity-preparation.ts), not universal
tessellation noise.
IFCREL* classes seen but never indexed as a relationship-graph edge.
Subset of byClass not recognised by the schema registry at all.
Every class seen, scanned-count descending.