fix: harden Windows state persistence retries

[ByBrawe]

Summary

Follow-up hardening for #5.

• Retry transient Windows rename failures before falling back to non-atomic copy/overwrite.
• Retry transient/partial JSON reads before treating session state as corrupt and returning an empty job list.
• Add deterministic regressions for simulated EPERM rename locks and partial JSON reads.
• Remove the duplicated Example progress.md heading introduced in fix: Windows-safe atomic session state writes (EPERM) #5.

Verification

• node --check src/index.js
• node --check scripts/comprehensive-test.mjs
• Targeted Windows state persistence regression tests passed locally.

This keeps the fallback from unnecessarily widening the non-atomic write window and prevents a transient partial read from looking like jobs: [].

[Copilot]

Pull request overview

This PR hardens Windows session state persistence by adding retries around transient filesystem and partial-read failures, reducing the chance that temporary locks or partial reads cause dropped jobs or unnecessary non-atomic writes.

Changes:

• Retry readState on transient/partial JSON reads before treating state as corrupt.
• Retry rename on transient Windows lock errors before falling back to non-atomic copyFile/overwrite in writeFileAtomically.
• Add deterministic comprehensive regressions for rename-lock retries and partial JSON read retries; remove duplicated README heading.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/index.js	Adds retry loops for transient state reads and for Windows rename failures before falling back to copy/overwrite.
scripts/comprehensive-test.mjs	Adds regressions simulating EPERM destination locks and partial JSON reads; exposes stateFile via the harness.
README.md	Removes a duplicated “Example progress.md” heading section.

Suppressed comments (1)

scripts/comprehensive-test.mjs:552

• This test can become flaky if loop-status (or related internals) reads the state file more than once; that would make partialReads exceed 3 even though the retry behavior is correct. Use a >= assertion so the test only enforces that retries happened before treating state as corrupt.

    await h.command("loop-status")
    assert.equal(partialReads, 3, "partial JSON should be retried before treating state as corrupt")
    assert.equal(h.records.toasts.at(-1)?.message, "1 loop job(s).")

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.