quest(wildcard): block the line on the #4279 datagram, origin, and benchmark findings

[kixelated]

Problem

#4279 (spread a prefix's pool across paths) auto-merged into this line before CI, with its last Codex round unanswered. Maintainer decisions from the 09-28 merged-PR audit: the findings below block the line from landing on main.

Approach

Three child quests on the wildcard line, all verified still present on the line branch:

• Datagrams behind SUBSCRIBE_START, the three Codex P1s:
  • An oversized first datagram commits SUBSCRIBE_START and drops earlier groups, in Rust and JS (r4117485530).
  • A JS start-sequence race lets group 5 go out after a START for 10 (r4117485532).
  • FETCH_OK admission lets a datagram through before its own SUBSCRIBE_START (r4117485533).
• JS origin granularity: JS records the reply-named origin per broadcast, latest wins; it must match Rust (r4113954489).
• Pool resolution benchmark: pool size x requested paths sweep, the Codex P2 (r4117485535).

The README gains a Quests section and drops its [S] size, since a README with children carries none. The edits sit in new files, the title line, and a new section above Related, so they merge cleanly with #4383.

Impact

• Quest files only. No API or wire change.

Alternatives

• One quest per datagram P1: rejected, all three sit on the same SUBSCRIBE_START ordering in the lite publisher and subscriber.

Follow-ups

• None beyond the new quests.

(Written by Opus 5.5)

🤖 Generated with Claude Code