# Conclusion under the revised Hi-GAL goal

The existing observations support follow-up of embedded-protostar candidates,
but do not currently identify a population-wide hidden-protostar fraction,
physical lifetime correction or a confirmed temporal stellar birth.
See the [verified core replays](../artifacts/core_review_20261006/verification.json).

E15 supplies the strongest positive core evidence: 12.8% of nearby catalogue
prestellar clumps have a 22/24-micron counterpart, with a SPICY candidate-YSO
rate of 9.3% compared with 0.38% for the remainder. This shows that a 70-micron
non-detection is insufficient to exclude embedded activity. It does not by
itself prove source ownership, an unbiased hidden fraction or a temporal birth.

The offset-controlled W2 estimators give approximately 9–25% point estimates,
with approximately 45–55% corrected-share scenarios. These are descriptive
sensitivity, not confidence bounds on the physical population. Column-correlated
reddening, neighbour projections, tracer efficiency and differing evolutionary
windows leave the net bias sign unknown. The reproduced very-red point estimate
is 0.1867762; making it reproducible does not repair those inference assumptions.

E27/E28 found no demonstrated CLIP increment beyond clump properties in the
external fragmentation/chemistry checks. E13 is qualified, E22/E26 superseded;
these tests neither establish physical hidden-protostar classification nor rule
out every possible future image method.
[E27 source](../artifacts/e27/incremental.json),
[E28 source](../artifacts/e28/incremental.json).

The [registered N20 pre-fit gate](N20_PREFIT_GATE_RESULTS_20261006.md) gives an
exact informative counterexample: distinct prevalences can yield identical
indicator observations and the same known class-1 response when class dependence
is allowed. Its parameters are mathematical examples, not Hi-GAL fractions.
A scientifically justified independence/dependence or class-response restriction
is needed before a fitted latent fraction is interpreted physically. Its
empirical coverage audit also stopped on repeated maser names with differing
missing distance entries; no complete joint support matrix or model was produced.
[Exact witness](../artifacts/n20_prefit_gate_20261006/exact_witness.json),
[terminal failure](../artifacts/n20_prefit_gate_20261006/terminal_failure.json).

The full 19,219-clump W2 turn-on search retained no birth events after E3:
all four were pre-existing YSO outbursts. The calculated upper limit is conditional
on detectable W2-bright events and transferring pilot injection efficiency;
it does not constrain W2-dark physical births.

The separate four unexplained E9 events are still unvetted. Three are downward
changes; one is brightening. All have significant pre-onset aperture emission,
which does not identify its emitter or establish the variable source’s ownership.
They fail to establish a new appearance under the current evidence, but aperture
emission alone does not physically refute every possible newly forming source
within that beam. G333.1049−0.5056 remains an unobserved follow-up target, not a
confirmed transition. No new source-specific measurement or human image review
was performed in this bounded cycle.
[Four-event evidence](../artifacts/n20_prefit_gate_20261006/temporal_evidence_recheck.json).

Stop before fitting N20 or propagating absolute lifetimes under the current
assumptions. Any distinct further unit must register a defensible identification
restriction or column-matched control, keep all missing/ambiguous objects,
and preserve source-specific temporal evidence requirements. No automatic model
successor, generic-star benchmark or new survey acquisition follows this failure.
E29–E160 remain an unmerged archive.

The local conclusion and updated core presentation are ready. Required external
Atlas restoration remains unpublished: the owner session was checked and still
has the E155 12-file readback permission prompt. Cancellation and restoration
await the existing explicit approval request. Neither rejection was bypassed.
The full revised goal is not marked achieved while publication remains unmet.