docs: D6 engine-side headless VT spike — recommend "not now" (WI-130)

[thedancingdeveloper]

Seventh PR of the terminal-attach-budget initiative (WI-121). Stacked on #672 (F5); all land on feat/terminal-attach-budget. Docs only, no runtime change.

What this is (D6 / WI-130)

The design-spike answer to: should the engine run a headless terminal emulator per session so a cold attach ships screen state (tmux-style) instead of byte history? Written up in a new docs/ROADMAP.md.

• Crate comparison — vt100 (best fit: minimal grid, smallest surface), alacritty_terminal / termwiz (heavier renderer/feature surface a snapshot server doesn't need).
• Grounded cost at 8 sessions — ~1.7–3.4 MB grid each (~14–27 MB total) on top of the raw ring, plus continuous VT parse CPU on the PTY hot path (a cost the raw-ring design pays only per attach). No live number taken — no stack is reachable from an agent session; H1's load session is the measurement path if revisited.
• Wire shape (snapshot-start kind:"screen" + a server escape stream), the server-VT-vs-xterm fidelity risk and how the H2 harness proves it, and the modest mobile win (F5 already gives a bounded one-write restore).

Recommendation: not now

F1–F5 (and F4) address both operator symptoms — switch re-stream and slow fresh open — by bounding and client-side serialization, without doubling engine memory or adding hot-path CPU. Revisit only if, once shipped and measured on prod, mobile cold attach is still bottlenecked on the tail, the client cache proves untenable, or a new server-authoritative-screen requirement appears. No new work items filed.

scripts/check_docs.py clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UtEFLJAhLiq4NfZCN2vuBb

[thedancingdeveloper]

Cleanup: superseded by the single promotion PR #676. All of this work is already integrated into feat/terminal-attach-budget and is reviewable there per-commit (each commit is one initiative item with a full message). Closing to keep the initiative to one PR; the commit is preserved.