[p135246]

Why

A RandomInfra... sampler exists for a head whose family has more than one member. A region head (ball, shell, tube, cylinder, cone, solid of revolution, ball hull, convex hull, quadric) has exactly one member, its vertex set: "Cardinality" is the constant 1, so RandomInfraBall[g, ball] returned the support with no randomness and RandomInfraBall[g, ball, 2] returned {}. Random points in a region are already RandomInfraPoint[g, region, n], the counterpart of RandomPoint[region, n].

What

• Breaking (2.8.0): RandomInfraBall, RandomInfraShell, RandomInfraTube, RandomInfraCylinder, RandomInfraCone, RandomInfraSolidOfRevolution, RandomInfraBallHull, RandomInfraConvexHull, RandomInfraQuadric removed without an alias (export, definitions, usage, reference page, guide chips, tutorial, SeeAlso entries, sampler-shape tests). RandomInfraPlane stays: the separating sets of a plane are a family.
• New reader InfraSupport[graph, head]: Keys @ InfraMeasurement[graph, head, "VertexDensity"], the sorted vertex list the members of any head pass through; a List of heads maps. Reference page, guide bullet, tests.
• The scene engine, the multiway explorer and the substrate highlighter resolve a region token through InfraSupport; bound values unchanged.
• Fix: the generic "Cardinality", "Length", "VertexDensity", "EdgeDensity" and InfraMemberQ clauses of InfraMeasurement apply only when the head has a "Graph" reading, and "Subgraph", "CountingMeasure", "RiemannianMeasure" only when it has a "VertexDensity". An unknown property of a head that defines its readings by UpValues (a sister paclet's heads) stays unevaluated and silent instead of emitting VertexList::graph and Table::iterb. Test with a density-only dummy head.
• CHANGELOG 2.8.0 with the Earlier | Now table; PacletInfo.wl 2.8.0.
• Rebuilt the 41 documentation notebooks whose Markdown source changed plus InfraSupport.nb; build_docs.wls: 0 dead links in sources and built pages, 291 Build.

Verification

• run_tests.wls: 2460/2460 before, 2464/2464 after (67 files, 0 failures).
• Probe: 275 exports, every export has a usage line and a docs/Symbols page and vice versa; Names["WolframInstituteInfraGeometryRandomInfra*"] contains none of the removed names; InfraSupport[g, InfraBall[c, 2]] === Keys @ InfraMeasurement[g, InfraBall[c, 2], "VertexDensity"] on GridGraph[{5, 5}]; the unknown-property calls leave $MessageList empty.

Migration

Earlier	Now
RandomInfraBall[g, InfraBall[c, r]], RandomInfraBall[g, c, r]	InfraSupport[g, InfraBall[c, r]]
RandomInfraBall[g, ball, All]	{InfraSupport[g, ball]}
RandomInfraBall[g, ball, n] meant as random points	RandomInfraPoint[g, ball, n]
the other region samplers, head or bare form	InfraSupport[g, Infra<Region>[...]]

[p135246]

Not merging this for now: the uncommitted work in the dev repo's InfraGeometry checkout goes further (removes FindInfraShell, InfraDensity and InfraSubgraph, renames the scene symbols) and its CHANGELOG folds in this PR's non-evaluation guards. Left open for that work to supersede or absorb. Note for whoever lands it: InfraCausality 2.1.0 calls InfraDensity in its kernel (the anchor rule) and documents InfraSubgraph as a set-head reader, so removing those breaks InfraCausality until it is ported.