Date: 2026-08-07 Status: quarantined; not an engine, API, data-pack, release, or roadmap dependency
The site-specific Jones/Paranal scattered-moonlight experiment is removed from
the active Moira physical-visibility product path. It is not part of
clear_sky_naked_eye_point_source_v1, the public Python or REST contract, the
packaged compatibility resources, or the Phase 7 release gate.
This decision does not invalidate the source audit or the research questions. It recognizes that the attempted Paranal composition was an optional experimental branch, not a prerequisite for releasing the already admitted general physical-visibility model.
moira-physical-heliacal-visibility 1.2.0 pack.1.3 compatibility contract,
environment variable, public type, public function, or response field is
shipped.The quarantined scope includes the experimental moonlight runtime module,
Jones component-pack and Paranal scenario builders and validators, failed and
interrupted response-geometry artifacts, experiment-only tests, the draft
1.3 compatibility contract, and the tentative public/facade/REST integration.
The pre-quarantine state is preserved outside the repository at:
C:\Users\nilad\Documents\Codex\2026-07-25\m\recovery\moira-physical-visibility-pre-paranal-quarantine-2026-08-07.zip
Archive SHA-256:
b2d0debc26126618c68c94959a188ae304547c1c150c97edb590d239abea6182.
The archive contains all 157 files that were modified or untracked before the quarantine cleanup. It is recovery evidence, not a release artifact.
This work must not silently return as a Phase 7 prerequisite. Reopening it requires a separately authorized research project with a new model identity, a bounded scientific question, an explicit stop condition, source-owned validation, and a release decision independent of the current physical model. Until then, historical Jones/Paranal checkpoints are research history only and must not be described as admitted product capability.
Phase 7 returns to the core release boundary: regenerate the public capability
and API inventories from current runtime truth, validate the exact 1.2.0
external pack, build and clean-install the wheel and sdist offline, rerun the
independent event oracles and regression gates, and record the resulting
release receipt. No site-specific moonlight scenario is required for closure.