[Jammy2211]

Summary

The shipped HST/MGE runner leaves production collection choices unresolved. Add a Pulse decision document with verified candidate FITS identities and plan-only resource bounds, and separate them from proposed hardware, software, input, load and correctness choices.

Link the document from the existing pilot task and update the campaign/next step. Keep the independently owned project correctness audit #417 with its current session. No new campaign, compute, witness execution or scientific decision.

Closes #58.

API Changes

None. Pulse prose, campaign next steps and generated board only. Existing last_checkin and measurement snapshots/timestamps are unchanged.

Scripts Changed

None.

Validation checklist (--auto run)

• Effective level: safe (docs, small); declared consequence: judge.
• Plan: issue docs: prepare HST MGE pilot collection choices #58, scoped by the user's explicit Pulse-side direction.
• Tests: 206 Pulse tests passed; ruff lint/format and offline board consistency passed.
• Smoke: not applicable; organ documentation only, no downstream script/API changes.
• Review: independent Sol review CLEAN at a4ea984. Verified pinned FITS identity, guarded plan-only behavior, board reproduction, preserved timestamps and approval boundaries.
• Heart: STALE, sole reason “release validation incomplete: no rehearsal for current source”; permits development PR, not release.
• Evidence: all three local candidate FITS bytes match pinned project e4266a13; SHA256, primary-header dimensions and driver plan-only bounds verified. No worker launched.

The 32-process / 16 CPU-hour overall resource ceiling remains a proposal; the shipped runner covers 31 workers / 15.5 hours and no independent correctness process. Data origin, production suitability, selected hardware/stack, full resolved settings, background-load method and correctness tolerances remain unresolved. No improvement/regression claim is made.

Tier: judge — human /prm. Generated by the PyAutoLabs agent workflow.