Broader unlabelled near/mid-infrared inputs retain uncertainty and uneven support

Can bounded unfiltered short-infrared inputs broaden stellar/dust representation development beyond the small far-infrared groups?

Register six disjoint half-open sky rectangles, native AllWISE W1-W4 and associated 2MASS J/H/Ks fields, four fit-input regions, separate D5 diagnostic and Orion application excluded from fit. Acquire only spatial predicates; six GETs, no retries/redirects/uploads/images, 8MiB responses, 12MiB stage and 5GiB reserve. Preserve missing/ambiguous rows and independently reconstruct native support before any fit.

Three adversarial input-boundary controls pass. All six public requests and native preparation return observed terminal exit 0. Independent streaming XML/Decimal replay reproduces 88,515 native cells including 16,586 nulls, 33,439 support cells, exact IDs beyond floating-point precision, units/types, bounds, roles, schemas and response provenance. Forty scientific bindings accepted.

1,967 distinct native catalogue sources retained across six regions. 1,685 have at least two measured WISE bands; 461 have a cautious region-conditional 2MASS association and 404 have at least two associated near-IR measurement pairs. Four fit regions contain 1,255 sources, 1,071 with two WISE bands and 279 with two cautiously associated near-IR bands. Raw responses total 2,966,469 bytes; native-stage footprint about 3.2 MiB. No model fit, reference readout, labels or physical classes adopted.

Next: Freeze these inputs, then register noise-aware missing/uncertain-association spectral phenotypes and ordinary-star/embedded readouts with matched raw-colour/brightness baselines and representative independent calibration/fresh transfer. Preserve all optical/Gaia, dust/PACS, singleton and physical membership/temporal birth-region scope. Full goal active; E121 zero, no Git rewrite. E156 dashboard update queued while separate E155 remote readback permission remains pending.

These are input-support counts, not verified stars, detector recall, population purity or discoveries. Near-IR support is strongly uneven: D3/D4 have only 30/14 supported pairs among 353/296 inputs. Association uniqueness is conditional on the queried collection, not full-catalogue physical identification; epochs/outside-field competitors remain. Hand-selected fields, catalogue detection thresholds, artifacts/blends/galaxies, unavailable errors and missing negative truth limit all-sky inference. Orion context is consumed development.