Post-hackathon roadmap: deferred items from v1.7 audit (T-7, T-12 Windows, Q-4/Q-6/Q-7/Q-8)

[sgwannabe]

Tracks items deferred during the v1.7 audit (umbrellas #29 ~ #37). Each deferral has a written rationale in tests/fixtures/ASSESSMENT.md and an explicit re-open trigger.

Items

From P3 umbrella #32

• T-7 — e2e mock-bootstrap harness (AskUserQuestion replay).
  • Deferred per tests/fixtures/ASSESSMENT.md — artifact-level fixtures cover the byte-stable contracts; mock harness has high maintenance cost during prompt iteration.
  • Re-open trigger: any Socratic regression that slips past the artifact fixtures and reaches a real run.
• T-12 — Windows CI matrix (macOS shipped in PR test(v1.7.0): Phase 3 Part C — T-2 / T-3 / T-8 + macOS CI (T-12) — closes umbrella #32 #54; Windows still outstanding).
  • Deferred — bash-on-windows requires WSL action + script rewrites. Most plugin scripts use Unix-only tooling (shasum, find -exec, mktemp).
  • Re-open trigger: real user reports failure on Windows.

From P8 umbrella #37

• Q-4 — Interview amend/retry path.
  • Deferred per ASSESSMENT.md — UX flow change with low hackathon-week value. Today's workaround: /pf:new --no-cache and redo the interview.
  • Re-open trigger: real user reports needing it.
• Q-6 — Multi-run spec import.
  • Deferred — power-user workflow, no demo-time value. Q-9 expected-socratic JSON files (now shipped) feed into Q-6 trivially when revisited.
  • Re-open trigger: first user asks about reusing a spec.
• Q-7 — Headless Chrome screenshots for gallery.
  • Deferred — cross-platform Chrome detection too risky during freeze week; F-7 `content-visibility: auto` already mitigates iframe perf cost on smaller profiles.
  • Re-open trigger: gallery feels sluggish on a 26-card max profile in real use.
• Q-8 — Adaptive/branching interview (interview-tree.json).
  • Deferred — depends on Q-9 data (now shipped, cheaper later) + couples to deferred T-7 mock harness for the umbrella DoD's deterministic-replay test.
  • Re-open trigger: Q-9 data shows clear branching axes that a flat interview can't capture, OR _filled_ratio is consistently <0.4 in real runs.

How to revive an item

1. Confirm the re-open trigger fired (reference incident, user feedback, or measurement).
2. Open a focused PR for that single item — do not attempt to bundle multiple deferrals.
3. Update tests/fixtures/ASSESSMENT.md to remove the relevant section.
4. Check this issue's box and add the shipping PR link.

Closes when

All 6 boxes ticked, with linking PRs.