fix(deep): recover accepted publication response loss

[mldangelo-oai]

Summary

Recover a Deep publication whose canonical writes succeeded but whose child-process response was lost. The host replays the exact acknowledged operation once through the existing publication checks. Failure to write the optional acknowledgment no longer turns a successfully delivered publication into an error.

Changes

• Record the staged draft and checkpoint only after validation and all canonical writes; require an exact match before replay and remove the acknowledgment during staging cleanup.
• Preserve generation, reducer selection, continuation and seal checks, both supported legacy workflows, worker identities and staged bytes.
• Distinguish Windows storage failures from path-integrity errors. Optional acknowledgment I/O may fail; canonical writes, path validation and v2 publication recording remain authoritative.

Testing

At 774b4d5690af94fcaefd155c869e5e600ca8f8a1:

• All five required portable checks, types and formatting passed; publication Python/MCP checks passed 27/26 cases.
• Source and installed-SDK publication matrices each passed 20 cases/328 assertions. Extracted and physically standalone plugin checks each passed 48 cases.
• Controls cover accepted-response loss, precommit failures, absent/mismatched acknowledgments, stale ownership, failed replay without a third attempt, canonical write errors and Windows storage/path-error branches. The unchanged parent fails the six Windows acknowledgment-I/O cases.
• Three fresh native source reviews and independent verification found no outstanding issues.

Broader suites ran at 91515fad38556d651d8387e6ecc13edde66062a6, before the test-only delete-on-close emulator correction; their runtime inputs are unchanged: Python 1,126 passed, 8 skipped, 109 subtests passed, 1 pre-existing fallback-test failure; MCP 72 passed; relevant SDK checks 98 passed. The fallback failure reproduces on the unchanged prerequisite. Final affected checks above include the corrected emulator.

Native platform CI passed at this exact head. The explicit Windows storage/path-error fault cases used Linux emulation; generic platform CI does not establish every native Win32 error-injection path. Earlier-head platform results are not inherited.

Risk and rollout

Stacked on #916, which must land first. No public command, flag, schema or workflow activation changes. Replay requires the exact acknowledgment and re-enters publication validation; process exit alone is not acceptance. An absent acknowledgment still prevents replay after response loss. Historical failed scans whose terminal reason was already lost are not reconstructed.

Public disclosure review

• No customer, partner, prospect, or user identities, data, or identifying details are included.
• No credentials, personal data, private source, scan findings, or nonpublic links or tickets are included.
• I reviewed the branch name, title, description, commits, changes, comments, logs, screenshots, attachments, and links for public disclosure.

[mldangelo-oai]

@codex review

Please review exact head f9aae3378e89a2214781f250aba22343a8a0f042 against prerequisite 4a5983f7aac613131d2ee8eec12a6b5170857136: accepted-operation identity, one replay at the existing owner, generation/selection/continuation checks, pre-acceptance failures and cleanup.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-16T15:23:05.050961Z	774b4d5	New commits
🔒 Security Review	✅ Completed 2026-09-16T15:25:40.054377Z	774b4d5	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. 🎉

Reviewed commit: f9aae3378e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f9aae3378e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".