probe: DO NOT MERGE — pre-dual-read tree on bun 1.4.2 (isolation for #70)

[ivndev001]

Throwaway isolation experiment for #70. Do not merge; I will close it as soon as CI reports.

Every historical run confounds two variables: every pre-dual-read commit ran bun 1.3.x and every post-dual-read commit ran 1.4.x. Today's pinned runs filled one missing cell — bun 1.3.14 on the post-dual-read tree, where a persona signs the run in before the first step hangs 30 s. This fills the other: the pre-dual-read tree (d8affa0b, which ran both stories in 1,962 ms and 808 ms on 1.3.14) on bun 1.4.2.

• If a story hangs here, the toolchain is the cause and the dual-read is exonerated.
• If both stay fast, the toolchain is not sufficient, and the code change is implicated — which matters, because that change is what release chore(release): 1.9.0 — publish the .dz/ dual-read #69 publishes.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

_{Comment @coderabbitai help to get the list of available commands.}

[ivndev001]

Experiment done, closing as planned. Result: on the pre-dual-read tree with bun 1.4.2, a persona whose credentials are wrong hangs at 30,004 ms while signs-in stays fast at 802 ms. Full 2x2 posted to #70.